The communications technology journal since 1924 1/2011 · PDF fileThe communications technology journal since 1924. 1/2011. Heterogeneous networks increasing cellular capacity . 4

The communications technology journal since 1924 1/2011

Heterogeneous networks – increasing cellular capacity 4

Mobile broadband second wave 10

Managing the growth of video over IP 16

Microwave capacity evolution 22

Analytics – the truth is in there 28

Operator-provided visual communication 34

E R I C S S O N R E V I E W • 1 2011

There is no doubt that the world is in the midst of a data boom. Society is changing: the shift is obvious. Increasing numbers of people are moving out of rural areas into ever-expanding urban communities, putting pressure on networks in hotspot areas, and consequently on operators to provide bigger, better and faster services. Heterogeneous networks have a significant role to play to support the envisioned numbers of subscribers in densely populated areas.

The population shift is coupled with changes in user behavior and expectations. People today use many connected devices and several sub-scriptions to perform similar tasks, in different scenarios; at home, while commuting, while traveling and at work. Expectations relating to perfor-mance and service are rising, and user-driven demands are shaping technology and how services are packaged.

Ubiquitous mobile broadband and the data traffic that comes with it pres-ent operators with both opportunities and challenges. On the one hand, hav-ing increasing volumes of data pres-ents opportunities to create new reve-nue streams. On the other hand, it also presents new challenges to keep users satisfied with services offered, per-formance and availability – all at the right price.

Analytics helps operators make sense of these changing patterns, to under-stand what services people want and when they want them and conse-quently reduce churn.

People want to be in control of their lives. By giving subscribers the information they need to manage their own accounts, they will use what they have paid for, pay for what they want and be willing to pay for ad hoc services. Intelligent packaging is key to the success of the second wave of mobile broadband.

Håkan ErikssonGroup CTO and President of Ericsson Silicon Valley

The media-content equation contains three parameters: content consumers, content providers and content delivery. All three elements affect the way technology will develop to create a universal, cost-beneficial media system and each one is driven by diverse motivational factors.

The world is changing, and so are people and their need for communica-tion. The massive reduction in the cost of consumer electronics over the past decade has made the prospect of inter-communicating devices a reality in some homes. Visual communication has traditionally been dominated by expensive and proprietary solutions. This area is a good example of where several key factors have now come together to make the mass-market delivery of this service possible, and operators have a key role to play in this process.

Ensuring the efficient flow of all this data is not just about good net-work design, intelligent nodes and a top-notch user experience. Sometimes, we need to improve the network’s foundations. You would not, for exam-ple, attempt to build a skyscraper on the foundations built for a two-story house. And the re-farming of lower frequency bands to support mobile broadband together with higher-order modulation in new higher-frequency bands make microwave a gigabit per second technology, supporting long-term capacity evolution.

Editorial

Some 18 months ago, data traffic in mobile networks surpassed voice. Today, mobile broadband together with smartphones and other connected devices are constantly driving global figures for data traffic volumes in mobile networks upward.

The data boom: opportunities and challenges

E R I C S S O N R E V I E W • 1 2011

The commun cat ons technology journal s nce 924 1/2011

Heterogeneous networks increasing cellular capac ty 4

Mobile broadband second wave 0

Managing the growth of video over IP 16

Microwave capacity evolu ion 22

Analytics the truth is in there 28

Operator provided visual communication 34

CONTENTS 1/2011

4 Heterogeneous networks – increasing cellular capacity Enhancing and densifying the HSPA/LTE multi-standard macro-cellular network are additional methods of meeting increasing traffic demands and fulfilling users’ high expectations for mobile broadband services. This article analyzes expansion strategies, covers important design choices for heterogeneous networks – focusing on transmission power differences – and explores how to coordinate low power nodes and macro base stations.

10 Mobile broadband second wave – differentiated offeringsThe speed and quality of a user’s mobile broadband connection will be at least as important a differentiator as the desirability of devices and subscription cost. Operators introducing new pricing models and smart connectivity to drive revenue growth need to offer their subscribers greater levels of control and flexibility while maximizing the use of network resources.

16 Managing the growth of video over IPTV and video consumer behavior is changing. While broadcast TV is still popular for news and live events, consumers are using a wider variety of platforms and different ways to view content. There is a very obvious shift towards media services that focus on the individual, are simple to use, and deliver on-demand content in a way that meets user expectations when it comes to quality. There is an increase in on-demand spending among consumers that is driven by the quality of the user experience, ease of access and good content. In short, users are willing to pay if the content, experience and price are right.

22 Microwave capacity evolution New RAN architectures such as HSPA-evolved, LTE and heterogeneous networks have led to an ever increasing demand for backhaul capacity. In this article, key emerging gigabit per second (Gbps) microwave technologies that help address this issue are explored.

28 Analytics – the truth is in thereFrom a framework perspective, several levels of processing are required before bits become intellect. The relationships between data (bits), information (data in context), knowledge (interpreted information) and wisdom (the ability to make the right decisions) form the basis of Ericsson’s holistic approach to cross-domain analytics, and how it allows many kinds of information to be interpreted using a common toolset. This implies that network information can be set into a user and service context, service behavior and acceptance can be related to network conditions, and consumer experience measurements can be related to all touch points.

34 Operator-provided visual communicationPrerequisites for the mass-market adoption of visual communication are good audio and video (AV) quality and optimum user experience at an appropriate cost. Trials have shown that users begin to appreciate visual-communication services once AV quality reaches a certain level. Although solutions that provide an acceptable level of quality are available, the equipment remains prohibitively expensive, preventing the large-scale adoption of services.

Ericsson Review unfolds and clarifies Ericsson’s technology and product strategy, showing customers how to make technology and solutions work. The journal is distributed to readers in more than 130 countries.

Address :Telefonaktiebolaget LM EricssonSE-164 83 Stockholm, SwedenPhone: +46 8 719 00 00Fax: +46 8 522 915 99

Internet:www ericsson.com/review

Subscriptions:Ericsson Review is distributed through Ericsson’s regional offices. If you wish to receive a copy, contact your local Ericsson company. Articles and additional material are pub ished on www ericsson.com/review. Here you can use the RSS feed to keep informed of the latest updates.

Distribution:Strömberg Distribution AB

Publisher: Håkan Eriksson

Editorial board: Håkan Andersson, Ulrika Bergström, Joakim Cerwall, Deirdre P. Doyle, Håkan Djuphammar, Dan Fahrman, Dag Helmfrid, Hans Hermansson, Jonas Högberg, Cenk Kirbas, Magnus Karlson, Filip Lindell, Kristin Lindqvist, Patrik Regårdh, Patrik Roseen, Krister Svanbro, Da ibor Turina

Editor: Deirdre P. Doyledeirdre.doyle@jgcommunication se

Chief subeditor: Birgitte van den Muyzenberg

Contributors: Tes in Seale, Michael Costello, Mark Tuite and Paul Eade.

Art director: Jan Sturestig

LayoutCarola Pilarz

Illustrations:Claes-Göran Andersson

Printer: Edita Västra Aros, Västerås

ISSN: 0014-0171

Volume: 89, 2011

Heterogeneous networks – increasing cellular capacity Heterogeneous networks are an attractive means of expanding mobile network capacity. A hetero geneous network is typically composed of multiple radio access technologies, architectures, transmission solutions, and base stations with varying transmission power levels.

dense urban environment to illustrate achievable performance. Box B details the reference system parameters..

Upgrading the radio access (HSPA or LTE) of existing sites enables very high user data rates and improved system capacity2, which can be fur-ther enhanced through the addition of more spectrum, more antennas, and advanced processing within and between nodes. Increasing capacity and data rates in this way is attractive as it alleviates the need for new sites. Figure 2 illustrates a reference system and the effect of each improvement approach on data volumes.

The reference system is a 10MHz HSPA system with an inter-site distance of 425m, achieving monthly data vol-umes per subscriber of 5.9GB in DL and 0.7GB in UL. By doubling the spectrum to 20MHz, data volumes for the DL approxi-mately double. Figure 2 shows data rates achievable at low load with 95 percent coverage probability. In the DL, data rates of tens of Mbps are achieved. In the UL, however, this data rate is signifi-cantly lower (a few 100kbps) and increas-ing spectrum does not improve the sit-uation. This condition is referred to as power limitation: data rates are limited by relatively low received power, which is due to large attenuation between ter-minal and base station caused by a com-bination of distance and challenging radio propagation (such as in indoor locations at the cell edge).

At some point, the capacity and/or data rates offered by the existing net-work with enhanced radio access will no longer be sufficient. If possible, den-sifying the macro network is an attrac-tive evolution path in these cases. In dense urban areas, networks exist with

Enhancing and densifying the HSPA/LTE multi-standard macro-cellular network are additional methods of meeting increasing traffic demands and fulfilling users’ high expectations for mobile broadband services. This article analyzes expansion strategies, covers important design choices for heteroge-neous networks – focusing on transmission power differences – and explores how to coordinate low power nodes and macro base stations.

Introduction Mobile broadband traffic has surpassed voice and is continuing to grow rapid-ly. This trend is set to continue, with global traffic figures expected to dou-ble annually over the next five years.1 By 2014, the average subscriber will con-sume about 1GB of data per month compared with today’s average figures that are around some hundred MB per month. This traffic growth, driven by new services and terminal capabilities,

is paralleled by user expectations for data rates similar to those of fixed broad-band. Actual figures per subscriber can vary greatly depending on geographical market, terminal type and subscription type; some users with mobile devices are already creating traffic in the order of gigabytes and predictions are estimat-ed to be several GB per month for some devices and certain user behavior. The mobile industry is, therefore, prepar-ing for data rates in the order of tens of Mbps for indoor use as well as outside and gigabyte traffic volumes.

Improving, densifying and complementing the macro network There are several approaches that can be taken to meet traffic and data rate demands (see Figure 1.) On a high lev-el, the key options to expand network capacity include:

improving the macro layer; densifying the macro layer; and complementing the macro layer with low power nodes, thereby creating a heterogeneous network.

These approaches are discussed here in more detail and use the example of a

BOX A Terms and abbreviations

3GPP 3rd Generation Partnership ProjectCDMA Code Division Multiple AccessCSG Closed Subscriber GroupDL downlinkdBm power ratio of the measured power in decibels referenced to one milliwattGSM Global System for Mobile CommunicationsHARQ Hybrid Automatic Repeat requestHeNB Home evolved Node BHSPA High-Speed Packet Access

ICIC Inter-Cell Interference CoordinationISD Inter-Site DistanceLPN low power nodeLTE Long Term EvolutionRNC Radio Network ControlllerRRU Remote Radio UnitUE User EquipmentUL uplinkWCDMA Wideband Code Division Multiple AccessWiFi Wireless Fidelity

SA R A L A N DST RÖM, A N DE R S F U RUSK Ã R, K L A S JOH A NSSON, L A ET I T I A FA LCON ET T I A N D F R E DR IC K RON E ST E DT

4

E R I C S S O N R E V I E W • 1 2011

Expanding mobile network capacity

inter-site distances down to 100-200m. Benefits of densification include: the number of sites is kept relatively low, and network performance is insensitive to traffic location. Figure 2 shows that by doubling the number of macro sites, DL capacity is doubled. The DL capacity per site remains more or less the same, since there are twice as many sites. UL capaci-ty is more than doubled as users become less power limited – better capacity per site, twice as many sites. A significant increase in UL data rates is therefore achieved.

Complementing the macro networks with low power nodes, such as micro and pico base stations, has been consid-ered a way to increase capacity for both GSM and CDMA systems for some time now (see references3-4). This approach offers very high capacity and data rates in areas covered by the low power nodes. Performance for users in the macro net-work improves if low power nodes can serve a significant number of hotspots and coverage holes. Deploying low pow-er nodes can be challenging, as perfor-mance depends on close proximity to where traffic is generated. In addition, due to the reduced range of low pow-er nodes, more of them are required. Overcoming these challenges requires proper design and integration of the low power nodes. Figure 2 shows results for the deployment of 12 pico base stations per macro site in traffic hotspots. This yields the same DL capacity increase as the previous two approaches (more spectrum and densification). However, a larger gain is achieved in the UL, which is a result of mitigating the power limita-tion. The resulting UL data rate improve-ment is greater than for the other two approaches.

The way to meet future capaci-ty demand is by combining all three approaches: improving the macro lay-er; densifying the macro layer; and adding pico nodes, as indicated by the last example in Figure 2. How these approaches are combined and in what order depends on the existing network, targeted volumes and data rates, as well as the technical and economical feasi-bility of each approach. Such a heteroge-neous network configuration, exploit-ing macro and low power nodes, can in principle support arbitrary data vol-umes and very high data rates.

Design options for heterogeneous networks Several aspects govern effective design of heterogeneous networks. From a demand perspective, traffic volumes, traffic location and target data rates are important. From a supply perspective the important aspects include radio environment, macro-cellular cover-age, site availability, backhaul transmis-sion, spectrum and integration with the existing macro network. Commercial aspects, such as technology competi-tion, business models, and marketing and pricing strategies must also be con-sidered. To summarize, Table 1 includes guidelines for some of the key design choices operators encounter.

Deployment aspects and choice of radio-access technologyHow to best complement the macro net-work depends on the network scenar-io. HSPA or LTE operating within the licensed spectrum should be used if the base station is deployed in a public area, or if coverage is important. If the

base station is well isolated from inter-ference and range is not crucial – if it is used in a private home, for example – then WiFi exploiting unlicensed or license-exempt bands is an attractive solution. For authentication, simple sign-on, and access to mobile operator services the WiFi access point should be connected to the mobile core network. A 3GPP-based HeNB provides little gain over WiFi in such scenarios. On the con-trary, HeNBs may create coverage holes or use spectrum that would otherwise be available for the macro layer.

In the network-design process, it is important to consider the business model. While a single operator often manages outdoor macro-cellular net-works in urban areas, indoor systems are often shared between operators (cf Distributed Antenna Systems). WiFi access points and similar smaller scale solutions are often user-deployed (by individuals, enterprises or a third par-ty), where access can be open for all sub-scribers or available for certain users only (Closed Subscriber Group

Combine toolsUnlimited performanceCombine toolsUnlimited performance

Densify macroExample: 425m –> 300m ISDDensify macroExample: 425m –> 300m ISD

Add low power nodesPico, RRU, relay, WiFi, femtoExample: 12picos (1W) per macro site

Add low power nodesPico, RRU, relay, WiFi, femtoExample: 12picos (1W) per macro site

Improve macroMore spectrum, more antennas,improved processing andcoordinationExample: 10MHz –> 20MHz

Improve macroMore spectrum, more antennas,improved processing andcoordinationExample: 10MHz –> 20MHz

FIGURE 1 Toolbox for higher capacities and data rates

5

E R I C S S O N R E V I E W • 1 2011

receive the radio transmission as well as a radio frequency processing unit. A central control unit that can collect baseband signals from several RRUs performs baseband signal processing and higher layer processing. The con-trol unit and its distributed antennas/RRUs must be directly connected via a low-latency and high-capacity inter-face. An optical fiber-based backhaul is suitable for RRU deployment and such solutions are increasingly being used in HSPA high-capacity networks.

Where RRUs are not applicable, a stand-alone base station can instead be connected to the radio network control-ler (RNC) for HSPA and the core network for LTE. In contrast to RRUs, stand-alone pico base stations have loose backhaul requirements and may, therefore, fit with networks that have a high-latency and low-capacity interface. Two exam-ples of cooperation schemes that can be applied to stand-alone picos are: soft handover in WCDMA Rel-99; and Inter-Cell Interference Coordination (ICIC) available in LTE release 8, which enabled simple interference manage-ment between base stations (picos and macros).

Additionally, a relay or repeater may be employed to improve coverage. The relay needs to communicate with the macro-cellular donor base stations, either inband or outband. If spectrum is available out-of-band relaying – using one band for the access link between terminal and relay and a separate band for the backhaul link between relay and donor base station – is the preferred approach.

Coexistence of macro and low power nodesOne of the basic issues with heteroge-neous networks is how to determine the spectrum to employ in each cell lay-er, and for each technology – HSPA and LTE. To attain the highest possible data rates, it is necessary to use at least as much bandwidth as the UE is capable of handling in each layer. UE capabil-ity in terms of frequency bands influ-ences spectrum possibilities: if capaci-ty (high traffic volume) is the driver or spectrum is scarce, then macro-cellu-lar carrier frequencies should be reused. However, such an approach requires good cell planning and radio resource

[CSG]). For public systems, particu-larly outdoors and in difficult radio envi-ronments, open access for all subscrib-ers is important so that users connect to the best base station. This explains the first rule of thumb in Table 1.

Local traffic hotspots can cover a wide area, such as an entire block and include several buildings. In such cases, deploy-ing an outdoor low power node that also covers indoor locations would be suit-able. If the existing macro-cellular grid is too sparse to meet the traffic demand and provide adequate indoor service, deploying outdoor low power nodes is a useful technique to achieve general coverage improvement. When traffic is concentrated to one specific indoor loca-tion, such as a shopping mall, indoor deployment is preferable.

Type of low power node and backhaul solution Backhaul transmission becomes more important as the number of nodes increase, in part because it will consti-tute a larger share of the total cost of

ownership, but also as the availability of fixed backhaul affects the feasible place-ments, installation costs, and time need-ed for site acquisition and installation. A low power node can be connected just to the core network or to the core net-work and other base stations. Each con-nection, also called backhaul link, may have different bandwidth and latency characteristics. The capacity of the back-haul link not only affects user through-put, but also the overall radio access net-work performance as a high-capacity backhaul allows for tighter coordina-tion between nodes.

There are several types of low power nodes that put different requirements on backhaul support. For networks where the backhaul has low-latency and high-capacity characteristics, deploy-ing remote radio units (RRUs) is the pre-ferred approach; otherwise stand-alone pico base stations is an option.

An RRU has the potential to improve overall network performance through tight coordination between nodes and usually comprises antennas to send/

Monthly volume per subscriber[GB]Monthly volume per subscriber[GB]

Data rate at 5th percentile[Mbps]

Data rate at 5th percentile[Mbps]

5050

4545

4040

3535

3030

2525

2020

1515

1010

55

00

3030

252543,043,0

20,720,7

5,15,1

9,29,29,49,4

18,718,7

12,512,5

9,29,2

11,711,7

5,95,92,82,8

12,212,2

6,86,8

1,11,10,70,7 0,300,30 0,460,46

3,03,01,31,3

21,721,7

2020

1515

1010

55

00Reference10MHz ISD 425m

Reference10MHz ISD 425m

Improve20MHz ISD 425m

Improve20MHz ISD 425m

Densify10MHz ISD 300m

Densify10MHz ISD 300m

Pico10MHz ISD 425m12pico

Pico10MHz ISD 425m12pico

Combined20MHz ISD 300m6pico

Combined20MHz ISD 300m6pico

FIGURE 2 DL and UL monthly data volumes and data rates at 5th percentile (95 percent coverage) supported by the different expansion strategies

FIGURE 2, KEY Performance evaluations Blue bars DL monthly data volumes

Light blue bars UL monthly data volumes

Blue markers UL 5th percentile data rates

Orange markers DL 5th percentile data rates

6

E R I C S S O N R E V I E W • 1 2011


management schemes to control inter-ference between cell layers. In particu-lar, mobility and control plane quality might be affected.

Our focus is on networks where macro-cellular carrier frequencies are reused throughout the network. By def-inition, a low power node has signifi-cantly lower transmission power than its surrounding macro base stations. Cell selection is typically based on DL received power, including the effects of the different base station transmission powers. This leads to an area surround-ing the low power node where the mac-ro base station is selected, but where the pathloss is lower towards the low pow-er node. In the UL direction, where the transmit power is the same, it would be better to be connected to the low power node also in this area. This is illustrat-ed in Figure 3. By increasing transmis-sion power, the cell size of low power nodes can be increased. However, doing so affects the cost and size of the node, which in turn limits site availability.

The range of the low power node can also be increased using a cell selection offset or handover thresholds that favor the selection of the low power node. This leads to the UL being received in the best node (the low power node) and offloads the macro to a greater extent. These benefits, however, come at the cost of higher DL interference for users on the edge of the low power node cell. Without further coordination of macro and low power nodes, there is a trade-off between DL and UL performance. In HSPA, soft handover functionality is useful to increase the UL low power node coverage and capacity.

Coordination potentialIn the situation just described signal strength is imbalanced. A highly prom-ising solution for improving perfor-mance in this case is based on cooper-ation between the macro and the low power nodes within its coverage area. For LTE DL, cooperation supports effi-cient offloading by extending the range of the low power cell. For UL, coopera-tion enables the macro base station to exploit UE signals received at pico base stations. This is favorable due to the pow-er-based cell selection (none or small cell selection offset) which creates a situa-tion where pico base stations are often

closer to the macro users than their serving macro base station and conse-quently, the pico base stations receive better-quality UE signals. There are dif-ferent flavors of cooperation schemes, such as coordinated scheduling, coor-dinated beamforming, as well as joint transmission and reception.

With coordinated beamforming, a reduction of the interference caused to a non-served user can be achieved by using an appropriate base station antenna pattern: a so-called beam. Due to their loose backhaul capacity and latency requirements, simplified coor-dinated beamforming schemes can be applied in a distributed pico-macro set-up. However, such schemes offer the highest potential in a centralized RRU

deployment that enables more elaborate optimization algorithms.

Joint transmission-based coopera-tion refers to simultaneous transmis-sion from different nodes to the same user. To achieve a coherent overlap of the signals at the receiver, the trans-mitters must be tightly synchronized in time and frequency. While transmit-ters can easily reach the required syn-chronization level in an RRU deploy-ment, additional synchronization equipment at each node (such as a GPS receiver) is needed in a distributed pico-macro setup. Therefore, joint transmis-sion is more easily applied in an RRU deployment.

There are diverse joint-recep-tion schemes in the UL, based on

BOX B Performance evaluations User behavior and traffic – 6,000 subscrib-ers per sq. km, 80 percent in indoor clusters (250 per sq. km), file transfer, 6 percent of daily traffic during busy hour

Deployment and propagation – Urban environ-ment, macro ISD 425m or 300m, clusters deployed with low power nodes in order of traffic volume

System – HSPA in 10 or 20MHz, antenna configu-ration DL 2×2, UL 1×2, macro power 46dBm, pico power 30dBm

Table 1: Rules of thumb for low power node deploymentDesign choice Decision criteria

Access Open access Closed subscriber group

Deployment conditions Operator deployed User deployed

DeploymentIndoor deploymentOutdoor deployment

Hotspot spread and position Large indoor hotspot Outdoor hotspot or many smaller indoor hotspots

Type of low power node RRUConventional pico Relay

Backhaul availability Fiber (P2P or WDM PON) Copper / fiber / microwave No backhaul

Frequency reuseReuse macro spectrumSeparate spectrum

Capacity need and access Capacity is driver Closed subscriber group

Power and cell selectionPower Biased cell selection

Hotspot area Cover the hotspot* Cover the hotspot* *value varies significantly

Signal from picostrongest

Signal from picostrongest

Signal from macrostrongest

Signal from macrostrongest

In the uplink, pathloss determines which basestation that receives the UE signal strongestIn the uplink, pathloss determines which basestation that receives the UE signal strongest

Signal from macro strongest in extended range,as the macro base station has higher transmitpower which compensates for the higher pathloss

Signal from macro strongest in extended range,as the macro base station has higher transmitpower which compensates for the higher pathloss

A cell section offset can beused to extend the rangeA cell section offset can beused to extend the range

FIGURE 3 Cell selection in a heterogeneous network

7

E R I C S S O N R E V I E W • 1 2011

more-or-less extensive informa-tion exchange between nodes. For WCDMA, the basic functionality of soft and softer handover represents a form of joint reception. In LTE, joint-reception schemes should preferably be applied in a network with a low-latency backhaul as the synchronous uplink HARQ has strict timing requirements.

Compared to a heterogeneous deploy-ment with stand-alone nodes, coor-dinated pico and RRU deployments enable straightforward optimization of joint-reception and joint-scheduling schemes. Examples of achievable gains are shown in Figure 4. Both data rates and capacity (achievable monthly vol-ume per subscriber) can be improved. The figure shows an increase in month-ly volume by a factor of 3.2 with a tight RRU deployment compared to an unco-ordinated heterogeneous deployment for a fixed required fifth percentile data rate of 0.5Mbps. Alternatively, for a fixed monthly volume per subscriber of 7GB, the improvement in the fifth percen-tile data rate reaches a factor of 12 with a tight RRU deployment.

ConclusionMobile-broadband traffic is increasing. In parallel, new applications are raising

expectations for higher data rates in UL and DL. Creating a heterogeneous net-work by introducing low power nodes is an attractive approach to meeting traf-fic demands and performance expecta-tions, particularly in situations where traffic is concentrated – in hotspots, or areas that cannot be suitably covered by the macro layer. By combining low power nodes with an improved and densified macro layer, very high traffic volumes and data rates can be support-ed. The nature of the existing network, as well as technical and economic con-siderations, will dictate which approach – improving the macro layer; densifying the macro layer; or adding pico nodes – or combination of approaches best meets volume and data-rate targets.

Low power nodes give high data rates locally and also offer benefits to mac-ro users by offloading and cooperat-ing with the macro layer. Tight integra-tion of low power nodes with the macro network provides gains over the unco-ordinated case through favorable com-bining of received signals and avoiding interference.

Coordination gain in the uplink[x factor]Coordination gain in the uplink[x factor]

1414

1212

1010

88

66

44

22

00

1.51.52.02.0

2.62.63.23.2

Pico deploy-ment loose coordination

Pico deploy-ment loose coordination

Data rate at 5th percentileData rate at 5th percentileMonthly volume per subscriberMonthly volume per subscriber

Pico deploy-ment tight coordination

Pico deploy-ment tight coordination

RRU deploy-ment loose coordination

RRU deploy-ment loose coordination

RRU deploy-ment tightcoordination

RRU deploy-ment tightcoordination

FIGURE 4 Coordination potential with two low power nodes in each macro cell cooperating with the macro base station, ISD 425m, 20MHz bandwidth, 1×2 antennas

8

E R I C S S O N R E V I E W • 1 2011


Sara Landström

is an experienced researcher at Ericsson Research in Luleå, Sweden. Her research area is Wireless Access

Networks and her current focus is LTE-based heterogeneous networks. She joined Ericsson in 2008 after receiving her Ph.D. in computer networking from Luleå University of Technology, Sweden.

Klas Johansson

is a system engineer at Ericsson’s WCDMA Systems Management in Kista, Sweden. He

received his Ph.D. in telecommunica-tions from KTH, Sweden in 2007 with a dissertation on cost-effective deploy-ment strategies for heterogeneous wireless networks. He joined Ericsson in 2008, and has led different activities related to HSPA evolution, including multi-carrier HSPA and the evolution of the enhanced uplink (EUL). He is currently a coordinator of hetero-geneous network activities for HSPA.

1. Mobile data traffic surpasses voice, http://www.ericsson.com/news/13969282. Next generation LTE, LTE-Advanced, S. Parkvall, A. Furuskär and E. Dahlman, Ericsson Review 2/2010

http://www.ericsson.com/news/101221_next_generation_lte_244218599_c3. M. Almgren, L. Bergström, M. Frodigh, K. Wallstedt, Channel Allocation and Power Settings in a Cellular

System with Macro and Micro Cells Using the Same Frequency Spectrum, Vehicular Technology Conference, 1996.

4. J. Shapira, Microcell Engineering in CDMA Cellular Networks, IEEE Transactions, Vehicular Technology Vol. 43, No.4, November 1994.

References

Laetitia Falconetti

is a research engineer in the Radio Protocols and Multimedia Technologies Group at Ericsson

Research, Aachen, Germany. Her research interests include interference management for LTE and energy- efficient mobile communications. Recently, her work has focused on suitable interference management techniques for heterogeneous LTE networks. She received an M.Sc. in electrical engineering from Karlsruhe University (KIT) in Germany in 2006. Before joining Ericsson in 2008, she was part of the 3GPP delegation of Rohde & Schwarz in Munich, Germany.

Fredric Kronestedt

joined Ericsson in 1994 to work on radio access network research. Since then, he has worked in

many different roles, including GSM system design and management. He is currently working at Business Unit Networks’, Development Unit Radio Systems and Technology, where he is a project manager working in close cooperation with network operators to analyze mobile broadband perfor-mance in real networks. Fredric holds an M.Sc. in electrical engineering from the Royal Institute of Technology (KTH), Sweden.

Anders Furuskär

is a principal researcher within the field of Wireless Access Networks at Ericsson Research. His current focus is HSPA and

LTE evolution to meet future data rate and traffic volume demands. Anders holds an M.Sc. and a Ph.D. from KTH, Sweden. He joined Ericsson in 1990.

9

E R I C S S O N R E V I E W • 1 2011

Mobile broadband second wave – differentiated offerings The speed and quality of a user’s mobile broadband connection will be at least as important a differentiator as the desirability of devices and subscription cost.

Addressing the second waveMobile broadband can be character-ized by three waves of revenue devel-opment for mobile operators, as shown in Figure 1.

The first wave – developments to date – was characterized by simple data sub-scription plans enabling people to con-nect their laptops, smartphones and net-books to the internet at the highest avail-able speeds. Essentially, there was one offering to fit all users, because simplicity was key to getting the market moving.

The second wave – the current focus in mature markets – moves this “one size fits all” approach along, to meet the needs of different subscribers and devices with differentiated services and traffic management – driving rev-enue growth through smart targeting. With differentiated connectivity servic-es, operators can provide offerings to fit all users using computers – as well as smartphones and tablets – while mak-ing the most of the mobile broadband business opportunity.

With differentiated offerings, oper-ators will be able to address more than just laptop, smartphone and netbook users, which will lead to the third wave of development of mobile broadband.

The third wave will be character-ized by everything being connected. Revenue will be driven by scaling to meet the needs of the more than 50 bil-lion connected devices envisioned. The proliferation of connected consumer electronics and industry-specific M2M devices will drive the need for a range of industry-specific cloud applications and non-traffic-based charging models.

3GPP and QoS – basis for a differentiation toolboxThe bearer, also known as PDP Context

Operators introducing new pricing models and smart connectivity to drive revenue growth need to offer their subscribers greater levels of control and flexibility while maximizing the use of network resources.

The differentiation toolbox helps operators differentiate their mobile broadband services by defining:

how to differentiate – by setting, for example, maximum bandwidth per user or per application, per-user admission priority, throughput per user, guaranteed bit rate per application, minimum bit rate per user, content optimization and content caching; and

when to differentiate – based on factors such as: fair usage policy, time of day, location, terminal type, detected service or subscriber interaction.

IntroductionIn the evolving broadband market, the speed and quality of the user’s connec-tion is as important in influencing sub-scribers in their choice of operator as the desirability of devices, the coolness of the apps and the cost of the subscription.

Besides introducing new network and device capabilities, operators want-ing to meet customers’ demands and expand their mobile broadband busi-ness need to:

maximize mobile broadband revenues by providing offerings with a balance between price and performance to fit all users; attract and retain customers by deliver-ing a differentiated user experience that meets the expectations of all mobile broadband users; and optimize network utilization.

Policy and charging control together with the QoS mechanisms available in the network are instrumental in meet-ing these needs.

M A RT I N L J U NGBE RG A N D A L DO BOL L E


2G second generation wireless telephone technology3G 3rd generation wireless telephone technology3GPP 3G Partnership ProjectARP Allocation/Retention PolicyARPU average revenue per userBSS Business Support SystemsDPI deep packet inspectionE2E end-to-endGBR guaranteed bit rateGW gatewayHSPA High-Speed Packet AccessIEEE Institute of Electrical and Electronics EngineersIMS IP Multimedia Subsystem

LTE Long Term EvolutionM2M machine-to-machineMBB mobile broadbandMBR maximum bit rateOSS Operational Support SystemsPDP Packet Data ProtocolQCI QoS Class IdentifierQoS quality of serviceRAN radio access networkSACC Service Aware Charging and ControlSPR subscriber profile repositoryTHP traffic handling priorityUMTS Universal Mobile Telecommunications SystemVoIP voice over IP

10

E R I C S S O N R E V I E W • 1 2011

Driving revenue growth

for 2G/3G networks, is a central element for supporting QoS in mobile networks. With 3GPP Release 8 for LTE radio access and Release 7 for 2G/3G radio access, QoS control has been further enhanced with the network-initiated QoS bearer- control paradigm. This concept provides network and service operators with a set of tools to further improve service and subscriber differentiation. Such tools are becoming increasingly impor-tant as operators are moving from sin-gle to multi-service offerings while the number of mobile broadband subscrib-ers and traffic volume per subscriber is rapidly increasing.

Providing subscriber and service differentiation, and positive and neg-ative traffic discrimination in accor-dance with operator and regulator policies, requires the involvement and integration of functionality in the entire network infrastructure, as shown in Figure 2, and specifically in:

Radio access and transport – including new capacity-enhancing techniques and congestion-aware scheduling of air interface resources; Packet core – including prioritization of data packets, negotiation or modifica-tion of QoS parameters, and content optimization and caching; and Business Support Systems (BSS) – including flexible charging options, tar-geted offerings, real-time credit control and greater customer control and inter-action, such as real-time notifications and self-service.

End-to-end QoS control in mobile broadband implies close interac-tion between the different enti-ties in the network, from the pol-icy controller (PCRF) and deep pack-et inspection (DPI) function, to the core nodes and radio and transport network. Only with direct, real-time delegation of policies from the poli-cy controller to the scheduler in the base station can operators monitor and enforce QoS dynamically according to congestion, location and device type.

E2E QoS control is an enabler for refined load-management strategies. It combines the capability of the DPI entity to identify OTT services and notify the PCRF, with the PCRF capa-bility to apply service and subscriber specific policies based on the QoS profile of the bearer.

Standard

One sizefits all

Demand-basedpricing/QoS

More than50 billion

connecteddevices

Price

Subscribers

Price

Subscribers

Connections

Time

Smart Scale

Establishing marketEstablishing market Differentiating servicesDifferentiating services Connecting everythingConnecting everything

First wave Second wave Third wave

FIGURE 1 The three waves of mobile broadband revenue development

Customer care Self-serviceportal

Provisioning Charging SPR

IMS

Policycontroller

GW DPI

BSS

Terminals RAN Packet core

Service layer

FIGURE 2 Network providing QoS and differentiation functionality

11

E R I C S S O N R E V I E W • 1 2011

Differentiation toolbox as a means to simplify the definition of use cases Operators have a range of mobile broad-band differentiation mechanisms at their disposal to implement E2E QoS and policy control to create a basis for subscriber and service differentiation. These mechanisms, together with operational and business support sys-tems (OSS and BSS), network planning and design, integration and optimiza-tion deliver the required granularity of control needed to create differentiated mobile broadband offerings.

Understanding the available differ-entiation capabilities is fundamental to empowering operators’ commercial and marketing strategies and objectives, so that operators can identify where invest-ments in mobile broadband capacity are needed, while avoiding over-dimension-ing and unnecessary expenditure. To facilitate this understanding, the differ-entiation mechanism can be organized into a “differentiation toolbox.”

The differentiation toolbox , as shown in Figure 3, includes negotiation pro-cedures and enforcement policies and mechanisms, which are supported in the different parts of the core, transmis-sion and radio networks. Implementing enforcement mechanisms across the

By combining the three elements of the differentiation toolbox, operators can define and realize service packag-es where actual user experience meets expectations.

How to differentiateA closer look at the “how to differenti-ate” shows how the MBB user experi-ence can be differentiated based on a combination of caps, priorities, guaran-tees and acceleration mechanisms.

Maximum bandwidth per user

This controls the peak data rate that users can achieve on the MBB connec-tion and is controlled via the maximum bit rate (MBR) parameter. With MBR per user enabled, peak rate is not deter-mined by network capability but rath-er by the subscription and the evaluat-ed QoS policies policies, as illustrated in Figure 4. Maximum bandwidth per application

Rather than applying limits to the entire traffic of a user, a peak data rate for spe-cific applications can be set instead, which the differentiation toolbox refers to as maximum bandwidth per applica-tion. This mechanism is implemented by limiting single IP traffic flows with-in a multi-service session and is often referred to as IP flow bandwidth man-agement or IP throttling.

Admission priority per user

When resources are limited, differen-tiation based on an admission priori-ty to access resources can be applied. This mechanism is regulated by the allocation and retention priority (ARP) parameter, which allocates a relative level of importance of the radio bearer compared with other radio bearers. It is used to allocate, retain or release radio resources in situations where the net-work approaches congestion. Different admission priorities can be set per subscriber or per service type – such as voice services versus non-real-time services.

Throughput per user

In congested situations, the differentia-tion provided by maximum bandwidth per user may not be sufficient to create a differentiated subscriber experience. To deliver the required experience to

entire network – radio, transport and core – is a key approach in achieving service differentiation, subscriber seg-mentation and load management. The differentiation toolbox highlights how operators can benefit by moving away from the reactive approach of purely limiting heavy users and take a more proactive, marketing-led approach to subscriber and service differentiation – where high usage is turned into a rev-enue opportunity rather than viewed as a problem. Two key elements of the toolbox define:

how to differentiate – by, for example, setting maximum bandwidth per user or per application, with per-user admission priority to the service, throughput per user at congestion, guaranteed bit rate per application, minimum bit rate per user, content optimization or content caching; and when to differentiate – based on factors such as fair usage policy, time of day, location, terminal type, detected service or subscriber interaction.

Differentiation control is the third ele-ment in the toolbox and is primarily related to the control of default and ded-icated bearers either at establishment or through modification. Certain policies and mechanisms can also be applied through IP flow management.

Maximumbandwidth per

user

Maximumbandwidth per

application

Per useradmission

priority

Throughput peruser

Guaranteedbit rate perapplication

Contentoptimization

Minimumbit rate per

user

Contentcaching

Fair usagepolicy

Time of the day

Location Terminal type

Detectedservice

Userinteraction

future policies

At bearerestablishmentstatic/dynamic

Throughdedicatedbearers

Through bearermodification

Through IP flow control

How to differentiate When to differentiate

Enforcement Negotiation

Differentiationcontrol

FIGURE 3 Differentiation toolbox for mobile broadband

12

E R I C S S O N R E V I E W • 1 2011


a premium package subscriber, a rela-tive priority to serve the different pack-age types (such as premium and stan-dard) can be introduced. This relative priority is controlled in the RAN and in transport by the traffic handling prior-ity (THP) parameter in UMTS networks, or the QoS Class Identifier (QCI) in LTE networks, which enables the association of a relative prioritization among inter-active bearers.

Figure 5 shows how changing THP (or similarly changing the QCI) for a heavy user gives the premium user a larger part of the available bandwidth in the case of cell congestion. Minimum bit rate and guaranteed bit rate

Differentiated minimum bit rate can be delivered by exploiting the sched-uling capabilities of the RAN. This parameter, although not explicitly signaled by the core network, can be configured and enforced in the RAN to ensure a differentiated experience. Unlike guaranteed bit rate, no admis-sion control is performed in the net-work for minimum bit rate. Instead the RAN will schedule traffic to sat-isfy minimum bit rate conditions as far as possible. If available resources are insufficient, the RAN still delivers a service, albeit at a lower bit rate.

When to differentiateWithin the “when to differentiate” sec-tion of the toolbox, conditions or events are listed that may trigger the applica-tion of a specific QoS.

For example, a fair usage policy con-dition may enable an evaluation and activation of a new QoS policy defini-tion when a certain accumulated usage threshold is reached. The thresholds are operator-configurable and can be used in different subscription packages. In contrast, operators may exploit other events or conditions, such as the loca-tion of the subscriber or the terminal type to apply an improved QoS when the user connects to the MBB service.

To deliver improved performance, some events, such as accessing a specif-ic site or a specific application, may trig-ger modification of a subscriber’s QoS profile. Users may also modify their QoS profile during the course of a day via a customer-care center or through a self-service portal.

QoS-differentiated offerings supported by Ericsson SACCBased on available network capabilities, Ericsson has analyzed, documented and verified a set of QoS use cases to rapid-ly support implementation of the most requested differentiation policies, com-bining the enforcement mechanism with the negotiation and trigger mech-anisms of the differentiation toolbox. A few examples of these use cases are listed below.

Tiered offerings (premium/standard/basic)Using maximum bandwidth, a differ-entiated subscription can be designed with higher speeds and higher prices. By adding admission priority, and traf-fic handling priority the more expen-sive subscriptions can be given higher priority, resulting in enhanced relative performance as the network becomes congested.

Fair usage policyUsers generating extensive traffic can be tricky; simply cutting-off or downgrad-ing a user to a very low maximum once usage quota has been exceeded might not be appreciated by premium users. At the same time, network resources have to be managed to provide good ser-vice levels in all conditions. A fair usage policy can be designed so that a user is

downgraded to lower admission prior-ity and traffic handling priority when usage quotas have been exceeded. This will secure performance for other users still on a higher priority and permit maximum performance to the heavy user as long as network resources are available.

As a commercial reference, one European operator offers tiered mobile broadband packages, with various QoS measures that kick in under certain circumstances. These include a fair usage policy that is enforced with different combinations of maximum bandwidth and de-prioritization of subscribers who have exceeded their monthly data volume allowance.

Services specific QoSSpecific services such as streaming audio/video or VoIP can be delivered with a better quality of experience through prioritized connectivity. When such services are detected using DPI or signaling, a minimum bit rate or GBR can provide the speeds needed for an improved user experience at times of congestion.

Another European operator has extensively tested prioritization for its mobile-TV application to ensure that subscribers paying for this ser-vice enjoy a good quality experi-ence, regardless of network load.

MBR[Mbps]

3

6

1

PremiumStandardBasic

FIGURE 4 Bandwidth management based on maximum bit rate

13

E R I C S S O N R E V I E W • 1 2011

Similarly, another European oper-ator prioritizes users of a certain smart-phone to preserve the user experi-ence of these high-ARPU subscribers whatever the network load.

On-demand and time of day (happy hour) A higher tier subscription can be given on demand for a shorter period, prefer-ably user activated, or at specific times of the day.

In commercial operation, one opera-tor in Southeast Asia enables its mobile broadband subscribers to check how much of their data allowance they have used through an online portal. Subscribers have the option to upgrade their package for a day, a week or a month – thus allowing users to return to their expected service level in the event of restricted throughput as a result of exceeding an allowance or cap.

ConclusionsThe challenges facing operators in the mobile broadband era include:

to stand out and attract users with the best possible quality of experience delivered from an excellent network; to sell mobile broadband services at price-performance points that suit all users, devices and services by introducing differentiation; and

to control costs while delivering a level of experience quality that consistently meets or exceeds user expectations.

The latest developments in 3GPP HSPA and LTE technologies provide the speeds and coverage expected from such an excellent network.

With a differentiation toolbox for mobile broadband services that includes a range of QoS, policy control and traf-fic management capabilities, operators can extend their business beyond the standard flat-rate service packages most commonly seen today.

This toolbox offers operators’ mobile broadband marketing and product development teams a way to create inno-vative services that are focused on open-ing up revenue opportunities, enhanc-ing customer relationships, building brand loyalty and, ultimately, boosting profitability.

At the same time, it helps operators control network load and cost.

Finally, the differentiation toolbox provides the operator with the features and functionality to achieve the desired subscriber experience.

Aggregatedbandwidth

Time

Without THPdifferentiationCell C

Requestedbandwidth Band-

width

Aggregatedload in cell

With THPdifferentiation

PremiumStandard

THP= 2 THP= 2

THP= 2 THP= 3

FIGURE 5 Traffic handling priority as cells become congested

14

E R I C S S O N R E V I E W • 1 2011


Aldo Bolle

is a system manager, within System and Technology, Product Development Unit Packet

Core, Business Unit Networks. He holds an M.Sc. in engineering physics, and a Ph.D. in electrical engineering from Chalmers University (Sweden). He joined Ericsson in 1997 as product manager for the microwave portfolio and has a background in research into optical fiber. Within Ericsson, he has been involved in the RAN requirements on the transport network and associated evolution towards IP, as well as the introduction of mobile broadband services. Recently, his focus has been on policy control and QoS enforcement mechanisms in 3GPP networks.

Martin Ljungberg

is a product manager for Mobile Broadband at Business Unit Networks and has a degree in engineering physics and

M.Sc. from Uppsala University (Sweden). He has the end-to-end product management responsibility for mobile broadband, including introduc-tion of LTE within Ericsson. He has been engaged in Ericsson’s mobile data and broadband activities since 1999 and has held a number of managerial positions with a focus on WCDMA and LTE, ranging from technical sales to product management and customer programs.

1. QoS Control in the 3GPP Evolved Packet System, Hannes Ekström, IEEE communications magazine, February 2009

2. Policy and Charging Control in the Evolved Packet System, José- Javier Pastor Balbás, Stefan Rommer and John Stenfelt, IEEE communications magazine, February 2009

3. 3GPP Tech. Spec. 23.203, “Policy and Charging Control Architecture.”

References

15

E R I C S S O N R E V I E W • 1 2011

16

E R I C S S O N R E V I E W • 1 2011

Managing the growth of video over IP The media-content equation contains three parameters: content consumers, content providers and content delivery. All three elements affect the way technology will develop to create a universal cost-beneficial media system and each one is driven by diverse motivational factors.

with the growth in managed and unmanaged content. MDN includes three core functions: content deliv-ery network (CDN), transparent inter-net caching (TIC), and service and performance enhancers (S&PEs). These functions enable fixed and mobile operators to:

create new revenue streams via whole-sale content distribution and delivery; differentiate through S&PE; and reduce opex and capex.

Handling the massive amount of OTT or unmanaged traffic is a major business challenge for network operators. Better compression techniques and improved connections have created a new business model in which countless video-based applications are generating substantial amounts of unmanaged traffic. Examples include YouTube, Apple TV, Netflix and applications from national broadcasters, such as BBC iPlayer, SVT Play, CBS Video and many others. Typically, these types of premium OTT services employ the capa-bilities of a CDN service provider. The CDN service provider distributes OTT content based on the agreement with the content owner across the internet as far as an exchange point, or in some cases further into an operator’s network when a cooperative agreement has been reached.

OTT applications can present net-work operators with several challenges:

rising backhaul transit costs; mounting last-mile bandwidth demands coupled with decreasing subscription revenues; and deeper penetration of CDN service providers’ caches in operator networks.

With internet traffic set to double every year over the next few years and the total market for CDN services estimat-ed to exceed USD 5 billion by 2015,

TV and video consumer behavior is changing. While broadcast TV is still popular for news and live events, consumers are using a wider variety of platforms and ways to view content. There is a very obvious shift toward media services that focus on the individ-ual, are simple to use, and deliver on-demand content in a way that meets users’ expectations when it comes to quality. On-demand spending is rising among consum-ers driven by the quality of the user experience, ease of access and good content. In short, users are willing to pay if the content, experience and price are right.

The change in the way people consume content – 50 percent of con-sumers now use internet TV on a weekly basis1 – has led to an increase in network data traffic. Mobile data traffic doubled in 2010 and is set to

double every year for the next three years2. The growth in connected devices and increased access to video indicate that both the behavior shift and data traffic growth will continue.

Content providers want to sell their products at competitive prices, continue to create revenue from advertising and deliver a personalized experience cater-ing for each consumer. They want to do this while maintaining brand value by ensuring that content is delivered according to user expectations and users get what they pay for.

Operators will incur increased costs to deliver increasingly large amounts of data efficiently while keeping custom-ers satisfied. There are various business models that operators can adopt to man-age the traffic growth and create new revenue streams by developing services.

Media Delivery NetworkEricsson introduced its Media Delivery Network (MDN) to help operators cope

IGNACIO M Á S, A L A N EVA NS, PAU L STA L L A R D, AYODE L E DA MOL A

Managing content for revenue


3GPP 3rd Generation Partnership ProjectATIS Alliance for Telecommunication Industry SolutionsBB broadbandBSS Business Support SystemCDC content delivery controlCDF content delivery functionCDL content delivery logicCDN content delivery networkDPI deep packet inspectionDSL digital subscriber lineDSLAM DSL access multiplexerETSI European Telecommunications Standards InstituteGGSN Gateway GPRS Support NodeGPRS General Packet Radio Service

GTP GPRS Tunneling ProtocolHLR home location registerHTTP Hypertext Transfer ProtocolIETF Internet Engineering Task ForceIPTV Internet Protocol TVMDN Media Delivery Network OSS Operational Support SystemOTT over-the-topP2P peer-to-peerQoS quality of serviceRNC radio network controllerS&PE service and performance enhancer SLA Service Level AgreementSTB set-top-boxTIC transparent internet cachingURL Uniform Resource Locator

17

E R I C S S O N R E V I E W • 1 2011

fixed and mobile network operators are investing in CDN solutions to efficiently manage the predicted traffic growth, and to explore new revenue opportunities by positioning themselves inside the digital-content value chain. Ericsson’s MDN is designed to assist network operators to overcome these challenges.

High-level MDN use cases There are four high-level use cases for Ericsson’s MDN. The first three are illus-trated in Figure 1, while the fourth – CDN federation – is shown in Figure 3.

Operator-managed In the first use case for MDN, operators efficiently deliver their own content to their subscribers. Here, the operator acquires content at a negoti-ated cost directly from content owners or aggregators, which can be viewed on a TV platform (IPTV for example), or on a variety of connected devices such as PCs, tablets, game consoles and mobile phones. Subscribers pay the operator for access to the content. MDN provides the operator with a unified delivery infrastructure capable of dis-tributing content across its network with minimum impact on the core, as well as providing the means to deliver that content to a wide variety of user devices. MDN replaces the traditional vertical silos that are usually deployed to provide these types of services, resulting in a simplified architecture and lower operating costs.

Wholesale CDNExternal content providers buy content-delivery capacity from the operator. MDN helps to minimize the traffic impact on the operator’s network, as well as providing a better quality of service (QoS) to end users, whose content will be delivered from edge nodes located close to their network access points. Content providers benefit from improved delivery of their content by using the operator’s optimized MDN-delivery capacity – saving them from having to purchase delivery capacity of their own – and from increased end-user retention due to the improved QoS.

MDN provides external content pro-viders with a comprehensive user inter-face through which they can manage and monitor content and access full

usage reporting. Through wholesale dis-tribution and delivery capabilities, MDN supports the operator to provide CDN services to multiple content owners in parallel with long-term agreements or on an event-by-event basis.

In addition to pure content distribu-tion and delivery, MDN allows the oper-ator to offer content management and adaptation services, such as ad insertion

and rights management. These value-added services provide further benefit to the content providers and additional revenue opportunities for the operator.

OTT service cachingOTT services are characterized by the lack of a business relationship between the content provider and the operator. In this use case, MDN’s

Content providers

Operator-managed Wholesale CDN OTT service caching

IPTV

Transitcosts

Reduced:backhaul costs andcore network traffic

FIGURE 1 Example use cases for Ericsson’s MDN

IP TV PC TVMobile

TV

WebOTT

MobileOTT

Fixed/mobile operator network

FIGURE 2 From unmanaged to managed data traffic

IP TV PC TVMobile

TV

WebOTT

MobileOTT

Fixed/mobile operator network

Ericsson MDN

18

E R I C S S O N R E V I E W • 1 2011

TIC capabilities allow the opera-tor to minimize the cost of deliver-ing OTT services. TIC is transparent to both the content provider and the consumer of the service. TIC helps the operator to significantly reduce peering and transit costs, as well as reducing the need to invest in internal network upgrades to support growing volumes of OTT traffic.

Some content providers are con-cerned that TIC adversely impacts their business by obscuring service usage and reducing advertising revenue. Although these fears can be allayed by modern TIC solutions that preserve application logic, some content service- providers intentionally undermine transparent caching by signaling content as un-cacheable or by using hashed URLs.

TIC provides a good short-term solution to minimize the impact of OTT traffic. The long-term strategy for most operators would be to create wholesale CDN agreements with the most prolific OTT content traffic generators.

CDN federationCDN federation – or CDN peering – allows independent CDNs to cooperate and deliver services across CDN bound-aries. One CDN might, for example, acquire content that is accessed by a subscriber located nearer to anoth-er CDN. CDN federation allows the CDN closest to the user to deliver the desired content and for the two CDNs to reconcile carriage payments and usage reporting. CDN federation has a number of important applications:

off-net distribution – operators can make content on their CDN available to subscribers outside their footprint. CDN federation supports global content delivery by enabling operators to buy capacity in peered CDNs; operator-to-operator peering – a collection or federation of CDNs appear as a single CDN. Internally, the CDNs aggregate usage information and cross-charge for carriage in much the same way as mobile operators do today with roaming. The federation model enables operator-based CDNs to compete for global content customers, who are likely to favor a single contract with one CDN with global reach over separate negotia-tions with each individual operator; and

Content provider

Content provider

Network ownerand CDN

Network ownerand CDN Network owner

and CDN

Internet and CDN service providers

Broadband subscribers

Broadband subscribersBroadband subscribers

FIGURE 3 CDN federation use case

Content deliverynetwork

Solicited contentand services

Unsolicited OTT

Reporting

Increase operatorshare of content

revenues

Decrease transitand peering costs

Service andperformanceenhancers

Network connectivity

Increase QoE

Transparentinternetcaching

FIGURE 4 The three components of Ericsson’s MDN solution


19

E R I C S S O N R E V I E W • 1 2011

global CDN-to-operator peering – global CDNs provide improved QoS to the edge of operators’ networks. Global CDNs federate with operator CDNs, purchas-ing capacity to deliver content deeper into the network, increasing QoS.

Although not a new concept, there are still no established standards for CDN federation. A significant amount of activity is ongoing to define a common standard within the inter-national standards organizations, including IETF, ETSI and ATIS. In the meantime, however, bespoke and ad-hoc CDN federations are forming. The emergence of a global standard will drive content providers to accept CDN federation and may fundamentally change the global CDN market.

MDN functional architectureAs illustrated in Figure 4, MDN includes three core functions: Content deliv-ery network; Service and performance enhancers; and Transparent internet caching. These functions enable fixed and mobile operators to reduce opex and capex by caching unmanaged OTT content, generating new revenue streams through wholesale content dis-tribution and delivery capabilities, and to differentiate themselves through enhanced service and performance.

Managed content: content delivery networkThe functional network architecture for a CDN integrated in an operator’s network is illustrated in Figure 5. Today, the control interfaces are not standardized and the main functional groups are as follows:

ingest function – receives new content and updates to existing content from the content owner. It contains procedures to store content in the CDN master storage and gets instructions from the ingest control function on how to handle updates and add new content; ingest control functions – authenticate content and content owners and respect business agreements with content own-ers. Ingest control can create, modify and delete metadata, check the consis-tency of metadata, deliver content to master storage and report usage statistics back to content providers; content metadata storage node – is a database that contains metadata

relating to content, such as: ownership, SLAs, ad insertion rules, distribution area rules, and locality and charging principles relative to the consumer; master storage – is where the master copy of content is stored. Content is then typically replicated to a set of dis-tributed content storage areas; content delivery control (CDC) – receives or intercepts content requests from a client device. Its decision logic determines what content to deliver and from which storage. The CDC controls the format and packaging, including personalization and ad insertion; content distribution logic (CDL) – is the brain of the system, determining where in the network each piece of content should be stored. CDL functions control where copies of content are stored, maintain consistency between master content and copies, initiate replication from one storage area to another and delete content from a storage area; content storage area – contains copies of content either from the master data storage node, a peer content storage node or content fetched on-the-fly; content delivery function (CDF) – pack-ages content in response to a client or

CDC request, acting on the decisions made by the CDC. CDF first fetches requested content and then delivers it to the requesting client, including protocol functions for HTTP, HTTP streaming and Flash; broadband network functions – performs L1/L4 network commands in an operator’s network and includes functions for transport, access, edge, authentication and security; subscriber and network data database – contains subscriber and operator net-work information, such as HLR, charging and user location, as well as network topology and fault status data; broadband client and access termina-tion – handles access and connection to the operator’s broadband network; and client function – the application used by, for example, standard web clients or IPTV clients.

Transparent internet cachingUnlike CDN architecture, TIC has no control-plane interface to the content owner. TIC is transparent – the content owner and the TIC do not interact with it and content owners are unaware of its existence. Due to its transparency,

Client

BB client BB NW functions

Content deliverycontrol

Ingestcontrol

Contentowner

Standardized services

Transport

Multi-access edgeAccess

Content distributionlogics

BB NW BB NW

CDL

CDC

Contentlocation

Ingestlogics

IngestBSS

Adinsert

Cachedelivery

Contentstorage

Masterstorage

IngestClientdelivery

Customized services

Sub andNW data

Contentmetadata

FIGURE 5 Functional architecture of MDN

20

E R I C S S O N R E V I E W • 1 2011

TIC has a thinner control layer than the CDN architecture, it has no master data storage, no ingest or ingest control functions, and no related BSS functions.

The user-plane interface between the TIC system and the content source is managed through broadband network functions. The TIC system captures requests and content by intercepting them in the normal user-data plane. TIC-specific functions are DPI and per-form content-management actions to ensure that captured and stored con-tent are synchronized with the original, maintaining accurate maps between URLs and content.

The TIC CDC is a light version of its CDN peer; TIC architecture lacks meta-data information and is less complex, as it only uses a subset of information.

Consequently, there is less need for control functions in the TIC system, and as most TIC control decisions can be taken locally in the network, all TIC functions can be naturally grouped into a single node type – the TIC node. Physically, this node will scale with: storage area, output delivery capacity

and DPI capacity to identify content and requests. The TIC node is local by nature and may be physically integrated into another network element.

Service and performance enhancersS&PEs are typically part of the TIC or CDN delivery node, such as: fast chan-nel change for IPTV, or compression and optimization functionalities for content adaptation in mobile networks.

CDN research activities For some time now, Ericsson’s CDN research group has been investigating content distribution technology with the aim of reducing bandwidth costs for operators. Several concepts have been explored: traffic redirection, caching, multi-protocol delivery, content migra-tion and redundancy elimination. These concepts have been applied to both fixed and mobile network architectures.

Traffic redirection addresses the impact of P2P file-sharing traffic on operator network bandwidth. The research group developed a method to reduce bandwidth utilization of

P2P traffic in fixed-broadband access networks using forced-forwarding techniques. This method involves local switching of P2P traffic at the DSLAM level instead of at the broadband-access router level. By eliminating the traffic tromboning effect, this method avoids wasting bandwidth.

The commercial deployment of IPTV systems presented a new research challenge: is P2P technology a viable method for redistributing video; in other words, utilizing the storage and computing resources of the end users’ STBs. This research developed the concept of multi-protocol CDN, by using several delivery protocols to bring content from the operator’s video head-end to the end user’s terminal. A time-shift TV prototype was developed based on P2P, unicast and multicast protocols. Multicast was used to populate network-based caches and caches in STBs during linear TV transmission. Once the linear TV distribution phase is complete, queries for previously transmitted TV programs result in a P2P request to the network caches and other STBs. This approach conserves network resources, as peers located closest in the network topology to the requesting STB serve the request.

Deploying caches in a network topol-ogy creates a new challenge: how to intelligently replicate content across these caches. A set of algorithms were developed to keep track of request rates for different content assets in a given section of the network and to decide whether to move or migrate content to caches closer to where most of the requests originate. These algorithms consider end-user quality of experience, number of content replicas in system, traffic load across network links and at the network peering point. The algorithms were implemented in a prototype that showed a reduction of overall bandwidth utilization both across the network and at the network peering point.

Terminal mobility and security architecture present additional chal-lenges to content distribution in 3GPP networks. For example, the presence of a GTP tunnel between the GGSN and the RNC limits the placement of caches in the mobile network. Session continuity after mobile handover can

1. Upload content

4. Searchfor content

5. Redirect

6. Redirect

7. Fetch

2. Decision to ingestbased on popularity or contract

3. Chooses initialplacement of content

8. Chooses optimalplacement of content

Contentprovider

Access andstorage provider

Cache

Cache

Allocator

Ingestor

Allocator

Dynamicnetworkmodel

FIGURE 6 Caching and content migration prototype


21

E R I C S S O N R E V I E W • 1 2011

be problem for video-streaming, where the cache is collocated with the RNC. To address this issue, a method based on redundancy elimination was pro-totyped. Two new functions – payload remover and payload inserter – together with a cache were introduced both in the core and at the access edge. These func-tions work as traffic compressors and decompressors, reducing the amount of traffic in the network link. This method enables seamless and stateless handling of mobile handover. The prototype showed that substantial traffic reduction in the mobile backhaul network is possible.

Further studies are ongoing in the area of caching at the RNC. Simulations have shown that cache equalization during off-peak hours enables the cache hit-rate of the system to equal that of a single large cache. Two modes for the RNC cache are being considered: transparent proxy mode enabling traffic reduction; and CDN server mode enabling hosting. Operators have varying needs and preferences and will be able to select the cache mode that is appropriate for them.

ConclusionsChanging user behavior and growing network traffic raise a business need for efficient media distribution and delivery. Ericsson’s Media Delivery Network addresses this need and helps operators reduce costs and increase rev-enue. The Ericsson solution differs from standard CDNs, as it includes media-specific capabilities as well as option-al network awareness. MDN combines capabilities such as wholesale content distribution and delivery, transparent internet caching and service enhancers that can be deployed as overlay architec-ture, embedded in network nodes.

In Ericsson’s view, both fixed and mobile operators will continue to be an important link in the digital-content value chain. Operators are best supported by a solution that enables them to manage both media traffic from the internet and traffic generated by OTT applications, while at the same time reduce transit costs and increase revenue by packaging their media-plane capabilities.

Ignacio Más

is a system architect at Ericsson Group Function Technology and is a senior specialist in networked

media architecture. He holds a Ph.D. in telecommunications from the Royal Institute of Technology (KTH) in Stock-holm, and an M.Sc. from both KTH and UPM (Madrid Polytechnic University). He joined Ericsson in 2005 and has worked in IETF standardization, IPTV and messaging architectures, as well as media-related activities for Ericsson Research. He is a member of the Ericsson System Architect Program (ESAP) and has research interests in quality of service, multimedia trans-port, signaling and network security, IPTV and, lately, cloud computing.

Paul Stallard

is vice president of systems management at Ericsson Television. He joined Ericsson in 1998,

and has worked in Research, Engineer-ing, Product Management and Solu-tions Architecture. In his current posi-tion he is responsible for coordinating the direction of Ericsson’s television portfolio. He holds a Ph.D. in electronic engineering from Bath University, UK, and prior to joining Ericsson was a lec-turer in computer science, specializing in computer architecture and network protocols.

1. Ericsson ConsumerLab, TV and Media Insights, http://www.ericsson.com/res/docs/2011/tv_and_media.pdf

2. Ericsson Annual Report 2010, http://www.ericsson.com/thecompany/investors/financial_reports/2010/ annual10/sites/default/files/Ericsson_AR_2010_EN.pdf

Alan Evans

Alan Evans is director of systems management at Ericsson Business Unit Multimedia. After joining

the company in 1998 he worked in Systems Integration, Systems Engi-neering, Product Management and Solutions Architecture. He is currently responsible for driving the portfolio and technology strategy of Ericsson’s Multimedia Business Unit. Alan holds a B.Sc. in electronic design engineering from the University of Huddersfield, UK.

Ayodele Damola

is an experienced research engineer at Ericsson Research. His research area is content

delivery networks and his current focus is content delivery and caching in mobile networks. He joined Ericsson in 2005 after receiving his M.Sc. in in-ternetworking from KTH, Stockholm.

References

22

E R I C S S O N R E V I E W • 1 2011

Microwave capacity evolution Microwave is a cost-efficient technology for flexible and rapid backhaul deployment to almost any location. It is the dominant backhaul media for mobile networks in the world today, and is expected to maintain this position during the evolution of mobile broadband.

in densely populated areas. A typical mobile-backhaul network has thou-sands of hops, and operators must be able to increase microwave capacity without having to change frequency planning and replace equipment across their entire network.

This article reviews current and future microwave solutions with the potential for Gbps backhaul capac-ity levels, for rural and urban areas. The solutions presented make best possible use of the advantages of micro-wave technology, including fast deploy-ment, flexibility and low total cost. The possibilities created by 112MHz chan-nels in the 42GHz band and by wider channel bandwidths in the 70/80GHz bands are explored. Finally, test-bed results are presented with record-breaking 1024QAM modulation, 5Gbps single-carrier radios, the world’s first line-of-sight (LoS) MIMO microwave demonstration and ultra-high spectral efficiencies offering Gbps transport capacity on a single 28MHz channel.

Breaking through the capacity barrierThis article defines capacity of a micro-wave link in two ways:

gross bit rate through the air; and line rate (the actual capacity available to users for data transportation).

Line rate is equal to the gross bit rate minus overhead data such as forward error-correction coding and frame over-head. Using efficient coding, native TDM and Ethernet transport, overhead data typically occupies 5 to10 percent of the gross bit rate.

A PDH microwave link in a 28MHz channel typically carries a line rate of 34Mbps. However, it is possible to reach Gbps bit rates through such a channel. Three methods can be used to increase the line rate:

improve the gross bit rate;

New RAN architectures such as HSPA-evolved, LTE and heterogeneous networks have led to an ever increasing demand for backhaul capacity. In this article, key emerging gigabit per second (Gbps) microwave technologies that help address this issue are explored.

BackgroundPoint-to-point microwave links are deployed using FDD with paired chan-nels, and typically hop-by-hop licensing. Frequency bands are governed by chan-nel plan recommendations from inter-national organizations such as CEPT/ECC and ITU-R, and span a wide range of frequencies from 6GHz to 86GHz – as shown in Figure 1. Today, the major-ity of point-to-point links are installed in the lower (blue) bands, whereas the higher (red) frequency bands offer wider bandwidths but are more vulnerable to precipitation. As such, the higher bands are more suitable for short hops up to a few kilometers in range.

National spectrum regulators control

the availability of frequency and chan-nel bandwidth; it is their responsibility to prioritize spectrum use among mil-itary, medical, space, automotive and fixed-service applications. Originally, the recommended channel bandwidths for point-to-point links were tailored for PDH/SDH transport networks based on a generic pattern of 3.5 or 2.5MHz chan-nel spacing. The most common chan-nels in Europe, for example, are 3.5, 7, 14 and 28MHz, with 56MHz channels available in some cases. Over the past few years, the rollout of mobile broad-band has fueled the trend toward using wider channels to enable greater capac-ity in the microwave link. In Europe, 112MHz channels have been intro-duced and use of the 3.5MHz channels is declining.

It is estimated that mobile-broad-band traffic will increase more than ten-fold during the next five years2. Many operators consider microwave to be a potential capacity bottleneck for future mobile backhaul due to its relatively narrow frequency channel and lack of spectrum in hotspot sites needed

JONA S H A NSRY D A N D JONA S E DSTA M

Microwave provides future backhaul capacity


3GPP 3rd Generation Partnership ProjectCEPT European Conference of Postal and Telecommunications AdministrationsDWDM dense wavelength division multiplexingECC Electronic Communications CommitteeFDD frequency-division duplexingGbE Gigabit EthernetGbps gigabits per secondHSPA High-Speed Packet AccessITU-R International Telecommunication Union – Radio-communicationLoS line-of-sight LTE Long-Term Evolution

Mbps megabits per secondMIMO multiple-input, multiple-outputm-QAM multi-level quadrature amplitude modulationPDH Plesiochronous Digital HierarchyQAM quadrature amplitude modulationRAN radio-access networkSDH synchronous digital hierarchySS-DP spatially separated dual-polarizedTDM time division multiplexingWCDMA Wideband Code Division Multiple AccessXPIC cross-polarization interference cancellation

BOX B In Figure 1, the blue bars indicate frequency bands used today, while red indicates frequency bands that are available, but currently not widely used for point-to-point microwave links.

23

E R I C S S O N R E V I E W • 1 2011

apply traffic optimization techniques; andreduce the radio overhead.

The gross bit rate can be increased by widening the used bandwidth and by improving spectral efficiency through increasing the number of bits transmit-ted per frequency bandwidth.

Traffic-optimization techniques remove redundant content from data streams before transmission. In an Ethernet stream, traffic optimization could, for example, include removing interframe gaps or compression of the transported data. Data compression efficiency is highly dependent on the type of data transported. If data is ran-domly distributed, as is the case for com-pressed and encrypted data streams, compression gain is limited or non- existent. Mobile-broadband payload traffic is generally compressed, and encryption is the default choice for WCDMA traffic and recommended by 3GPP for LTE.

As radio overhead is considered to be low, and traffic optimization is generally applicable to any type of capacity-limited transport medium, this article focuses on methods for increasing the gross bit rate of a microwave link. Presented below are methods for:

increasing operational bandwidth by using new frequency bands or multi- carrier channel aggregation; andimproving spectral efficiency in a micro-wave link with emphasis on high-order modulation techniques and LoS MIMO.

The impact on reach and availability is discussed, as well as how adaptive modulation techniques can be used to extend the reach of high-capacity micro-wave links while guaranteeing a mini-mum bit rate with high availability.

Increasing the gross bit rate Channel bandwidth limits the num-ber of symbols per second that can be transmitted on a microwave carrier. Using advanced signal-processing tech-niques, the symbol rate may, in mod-ern microwave systems, reach up to 0.9 times the channel bandwidth without violating spectrum masks. This implies that about 50Mbaud may be used for sig-naling in a 56MHz channel bandwidth. With m-QAM modulation, log2(m) bits may be coded onto each symbol. Thus,

using a 56MHz channel and 1024QAM modulation – 10 bits per symbol – a gross bit rate of 500Mbps may be trans-mitted on a single carrier.

Receiver sensitivity is reduced by about 3dB for every extra bit coded onto the signal. Consequently, an increase from 2 bits per symbol (4QAM) to 10 bits per symbol (1024QAM) results in a loss in total system gain of at least 24dB (see Equation 1). In practice, the reduc-tion in system gain will be even greater because the increased order of modu-lation places more extensive require-ments on the linearity of the radio, thus limiting the transmitter output power. Initial capacity gain is inexpensive; with an increase from 2QAM to 4QAM, the gross bit rate is doubled at the expense of 3dB loss in system margin. However, for higher-order modulation, capacity increase comes at a high cost for sys-tem gain. For example, an increase from 512QAM to 1024QAM results in a capac-ity increase of just 11 percent.

Adaptive modulationMicrowave links need to be designed with a sufficient fading margin to cater for signal deterioration caused by rain or multipath propagation, for exam-ple. In this way, high-priority traffic – such as voice – can be transmitted with maintained availability during periods of heavy rain. Because heavy rain condi-tions occur infrequently, the additional fading margin can be used to transport data by increasing the order of modu-lation. In the event of rain, path loss will increase and the link will switch to a lower-modulation format, saving system gain and preventing the link from being interrupted. This method

of adapting the modulation format is referred to as adaptive modulation.

When channel attenuation is high, traffic throughput for low-priority services can be reduced for short periods of time while maintaining normal availability for high-priority services. A link, originally designed for 4QAM, can in most cases support up to 1024QAM modulation 99 to 99.9 percent of the time, while maintaining telecom-grade availability (99.999 percent) for high-priority services supported by the 4QAM modulation.

In practice, adaptive modulation is a prerequisite for highly available, high-order modulation point-to-point micro-wave links over long hop lengths – tens of kilometers. However, to reach Gbps transport rates on 56MHz channels and below, multi-carrier solutions are needed. With wider channels, such as 112MHz and above, single-carrier solu-tions with Gbps capacity are possible.

Radio-link bondingThe latest RAN technologies – HSPA-evolved and LTE – enable higher gross bit rates through carrier aggregation. Microwave links for trunk (long-haul) networks have used this capability for a long time, and today it is also avail-able for microwave links in the access network. This method of transporting traffic evenly across two or more micro-wave carriers is referred to as radio-link bonding.

Radio-link bonding also supports an efficient link-protection scheme, where both links can be used for trans-port. If one of the links fails, the second radio will maintain support for high-priority services. This is known as

0 10 20 30 40 50 60 70 80 90

Frequency [GHz]

FIGURE 1 Microwave frequency bands recommended for fixed services by ITU-R and CEPT/ECC1

0

2

4

6

8

10

12

14

16

18

20

System gain (dB)

Max hop length (km)

110 120 130 140 150 160 170 180 190 200

7GHz10GHz23GHz38GHz42GHz72GHz

FIGURE 2 Maximum hop length limited by path attenuation for 70mm/h rain versus system gain and carrier frequency

24

E R I C S S O N R E V I E W • 1 2011

2+0 protection, as opposed to 1+1 protection, where one link is used to transport and the other remains in standby mode for backup purposes. The same protection scheme could be applied to intense-rain regions that lack rain-tolerant channels. Capacity can be increased by combining a low- frequency rain-tolerant microwave link with a high-freqency link with reduced rain tolerance. The low-frequency link will guarantee high-priority services while the high-frequency link, with reduced availability, will support lower- priority services.

Equation 1System gain = Ptx - PRxth + Gant1 + Gant2 where: Ptx is the transmitted output power; PRxth is the receiver threshold; and Gant1 and Gant2 are the antenna gains of radios 1 and 2, respectively.

New frequency bandsThe lack of spectra supporting wide- channel bandwidths has been iden-tified as a potential bottleneck for microwave backhaul. Many national regulators have recently adopted chan-nel plans that allow for bandwidths of up

to 112MHz in bands below 40GHz. These bands were originally made available at a time when there was limited need for wide bandwidths, and as a result, they are populated mostly with narrow channels. Since the rollout of mobile broadband, many of the narrowest channels have been abandoned because they are unsuitable for data traffic. This has given spectrum administrators the opportunity to re-farm these bands with wider channels. An additional possibili-ty is to open new, previously unused fre-quency bands such as the 42GHz band.

Box C details the CEPT/ECC recom-mendation for how the 42GHz and 70/80GHz bands can deliver 112MHz and 250MHz channel bandwidths. Using a single 112MHz channel, at 42GHz, it is possible to reach a gross bit rate of 1Gbps using 1024QAM modula-tion. In the 70/80GHz band, it is possible to aggregate neighboring channels to create wider ones. The Ericsson PT 6010 70/80GHz radio uses four 250MHz chan-nels (1GHz bandwidth), to support line rates of 1GbE.

At the Mobile World Congress (MWC) in Barcelona in February 2011, Ericsson demonstrated a full-duplex 5Gbps radio

using 2.5Gbaud and 4QAM modulation. As mentioned, higher-frequency

bands (30GHz and above) are more sensitive to rain attenuation5, which consequently limits the maximum hop length for these bands in comparison with lower ones. Typical hop lengths range from 2km to 4km in most climate zones for carrier frequencies between 30GHz and 42GHz with five-nines (99 999 percent) availability. To achieve the same availability, the hop length for a 70/80GHz radio is somewhere between 1km and 2km, with appropri-ate hardware. Figure 2 shows calcu-lated path-attenuated maximum hop length versus carrier frequency and typ-ical system gain at a rain rate of 70mm/h – representing a typical European climate zone.

Path attenuation in Figure 2 is cal-culated according to recommendation ITU-R P.530-136. The rain-attenuated distance is greater for lower frequen-cies. Other effects, such as multi-path propagation and the curvature of the earth, may become limiting factors for hop lengths above 20km. With a typi-cal system gain of 160-180dB, a realistic hop length with five-nines availability is 3km for a 42GHz link and 2km for a 70/80GHz link.

Example 1A user has access to two separate single-polarized channels between two nodes:

a 28MHz channel at 15GHz; and a 28MHz channel at 38GHz.

Each carrier is modulated with 1024QAM. Individually, each 28MHz carrier may transport a gross bit rate of 250Mbps. When aggregated, the two carriers can transport 500Mbps in a total bandwidth of 56MHz (28+28 MHz).

Spectrally efficient microwave linksA legacy microwave link on a 28MHz channel typically has a gross bit rate of 37Mbps, which corresponds to a spec-tral efficiency of 1.3bps/Hz. Although this may appear to be spectrally effi-cient, Figure 3 shows how spectral efficiency has increased by a factor of 30 in the past 10 years. In 2000, by spec-trally shaping the signal at the trans-mitter, the symbol rate was increased from ~18.5Mbaud to ~25Mbaud on a


BOX C New frequency bands channel bandwidths CEPT/ECC recommendation (01)04-2010 divides the 42GHz band, allocated from 40.5GHz to 43.5GHz, into: 12 x 112MHz; 25 x 56MHz; 50 x 28MHz paired channels; and a number of 3.5MHz, 7MHz and 14MHz channels3.

The 70/80GHz band is allocated from: 71-76GHz and from 81-86GHz with a total of 20 paired 250MHz channels that can also be aggregated4.

25

E R I C S S O N R E V I E W • 1 2011

28MHz channel, corresponding to a spectral efficiency of 1.8bps/Hz. Since then, the order of modulation has been increased continuously, with 512QAM commercially available since mid-2010. In addition, a rate of 1024QAM using commercial radios has been demon-strated in labs.

Support for higher-order modulation is planned, but moving to higher-order modulation increases the requirements on the radio transmitter and receiver in terms of linearity and phase noise. Pre-distorting the signal at the transmit-ter using digital signal processing can compensate for non-linearity, and phase noise can be reduced through careful design of the radio and modem. Moving from 4QAM to 1024QAM corresponds to a fivefold increase in capacity, which in the above example would correspond to a gross bit rate of 250Mbps (8 9bps/Hz).

Increasing the gross bit rate to Gbps levels requires multiple carriers for 56MHz and narrower channels. The next section includes a discussion about two multi-carrier techniques – polarization multiplexing and spatial multiplexing – that enable multiple carriers to share the same channel, further increasing spectral efficiency.

Polarization multiplexingPolarization multiplexing is a method for doubling spectral efficiency on a single channel. This method has been commer-cially available for point-to-point micro-wave links in the access network since mid-2000. It involves two single-carrier radios transmitting on the same frequen-cy channel but with orthogonal polariza-tions (horizontal and vertical). Because the radios share the same carrier frequen-cy, they can also share the same antenna.

Ideally, the two polarizations are completely isolated from each other. However, in practice, a small portion of the signal in one polarization will leak into the other. This can occur due to rotational misalignment between the antennas. In practice, it is difficult to achieve better isolation between the two polarizations than 25dB. Certain weath-er conditions, such as heavy rain, may further reduce the level of isolation that can reasonably be achieved. However, by sharing the received signal between the two modems, it is possible to cancel the interfering polarization using digital

signal processing. This technique is referred to as cross-polarization interfer-ence cancellation (XPIC).

Example 2 A user has access to a single 56MHz channel between two nodes. Using 1024QAM and 50Mbaud, a gross bit rate of 500Mbps can be transported per carrier – or a gross bit rate of 1Gbps using polarization multiplexing.

Line-of-sight MIMOMIMO is a well-known technology for increasing spectral efficiency in WiFi and RANs. An NxN MIMO system com-prises N transmitters and N receivers with the potential to simultaneous-ly transmit N independent signals. For example, a 2×2 MIMO system contains two transmitters and two receivers, and can transport two independent

signals, thus doubling the link’s capac-ity. The basic principle of MIMO is that a signal will use different paths between transmitters and receivers. In a 2×2 MIMO system, there are two possible paths between one transmitter and two receivers, as shown in Figure 4. The interfering signal can be cancelled if the difference in propagation between the two paths permits the two received signals to be orthogonal to each other at the receiver modems7-8. For a 2×2 sys-tem, this corresponds to a relative phase difference of 90 degrees. In convention-al MIMO systems, the difference in path is achieved through reflexes in the envi-ronment. For microwave links, it is not possible to take advantage of objects in the environment because these links, by definition, are operated in LoS mode with highly directional antennas. In contrast, because the carrier

4QAM 1024QAM

Spectral shaping:~1.5 x capacity per channel28MHz

28MHz

m-QAM modulation:log2m x capacity per channel

Polarizationmultiplexing (XPIC)2 x capacity per channel

N x N LoS MIMON x capacity per channel

FIGURE 3 Evolution of spectral efficiency for microwave links

0.001

0.01

0.1

1

10

100

Channel bandwidth [MHz]

Total line rate [Gbps]

64QAM - 5.4bps/Hz256QAM - 7.2bps/Hz1024QAM - 9bps/Hz1024QAM and XPIC- 18bps/Hz1024QAM + XPIC + 2x2 LoS MIMO - 36bps/Hz

7 14 28 56 112 250 500 1,000

FIGURE 5 Gross bit rates for a single-channel microwave link using different channel bandwidths, levels of modulation, and carrier multiplexing techniques

BOX D Figure 6 shows an experimental outdoor LoS MIMO setup at Ericsson’s test site in Mölndal, Sweden. Shown are four 32GHz radios mounted in a 4x4 SS-DP LoS MIMO configuration. The hop length is 1.3km and antenna separa-tion distance is 2.5m.

26

E R I C S S O N R E V I E W • 1 2011

where: f is the carrier frequency; c is the speed of light in air; and D is the hop length.

Figure 5 shows theoretical gross bit rates for a single microwave channel using the methods described above. With single-carrier radios it is possi-ble to reach 1Gbps gross bit rates for 112MHz channels. Using dual-carrier configurations (XPIC or LoS MIMO) it is possible to reach Gbps throughput on a 56MHz channel. Using quad-carrier configurations by combining XPIC and

frequencies for microwave links are high, it is possible to design a 2×2 MIMO channel with a phase difference of 90 degrees between short and long paths by spatially separating the radio antennas. This is commonly referred to as a LoS MIMO system. A geometric calculation indicates that the product of the two optimum antenna separa-tion distances d1 and d2 should fulfill Equation 2.

Equation 2 d1d2 = Dc/2f

LoS MIMO, it is possible to attain a Gbps gross bit rate on a single 28MHz chan-nel- We refer to this configuration as 4x4 spatially separated dual-polarized (SS-DP) LoS MIMO.

As shown in Figure 5, extrapolating the spectral efficiency of 36bps/Hz to a 112MHz channel results in a gross bit rate of 3.6Gbps on a single channel. Taking it one step further and apply-ing the same spectral efficiency to a 70/80GHz, 250MHz or 1,000MHz channel, we should be able to trans-port 8.1Gbps and 32Gbps respectively. These capacities are similar to trans-port levels on single channels in optical DWDM systems – indeed these trans-mission rates equal those achieved in current optical metro and core transport networks.

Demo at MWC 2011Ericsson demonstrated 1Gbps through-put in a 28MHz channel for the first time at the Mobile World Congress in Barcelona in February 2011. By combin-ing a 2x2 LoS MIMO system with polar-ization multiplexing, four commercial MINI-LINK radios at 10GHz were able to transmit four signals independent-ly of each other. Using a single 28MHz channel and 512QAM modulation, a line rate corresponding to 1GbE was demonstrated with a spectral efficien-cy of 32.1bps/Hz. This is, to the best of our knowledge, the highest reported spectral efficiency achieved over a microwave link. Increasing the order of modulation to 1024QAM would result in a spectral efficiency of 36bps/Hz. A MINI-LINK system operating at a 28MHz channel bandwidth, 1024QAM modu-lation and 46dBi antenna gain would obtain a system gain of 160dB or high-er for a carrier frequency below 40GHz. As shown in Figure 2, a 160dB system gain would – depending on the carrier frequency used – support hop lengths of tens of kilometers with Gbps capaci-ty levels under 70mm/h rain conditions. This corresponds to five-nines availabili-ty in a European climate zone and shows that it is possible to support Gbps bit rates across hop lengths of the order of tens of kilometers in a 28MHz channel with telecom-grade availability.

Summary This article provides an overview of


+

90°90°

d1 d2

D

D

D+∆

D+∆

+=

=

-90°

-90°

FIGURE 4 2x2 LoS MIMO basic principle

27

E R I C S S O N R E V I E W • 1 2011

microwave technologies that support the long-term capacity evolution of mobile broadband backhaul networks. Very high order modulation, adaptive modulation, radio-link bonding, polar-ization multiplexing and LoS MIMO are technologies that enable capacities in the order of several Gbps for microwave backhaul using commonly available frequency bands (6 to 40GHz). Applying these technologies to the wider chan-nels of the newly available 42GHz and 70/80GHz frequency bands makes it possible to achieve backhaul capaci-ties approaching 10Gbps and 40Gbps. Microwave is thus truly a fiber-through-the-air technology capable of support-ing rapid mobile broadband deploy-ment, and will remain an attractive and competitive choice for backhaul.

Jens Albrektsson, Patrik Bohlin, Antonio Carugati, Mats Cedheim, Donato Centrone, Jingjing Chen, Tomas Danielson, Carmelo Decanis, Maria Edberg, Thomas Emanuelsson, Pasquale Esposito, Duccio Gerli, Kåre Gustafsson, Magnus Gustafsson, Ola Gustafsson, Björn Gäfvert, Andreas Hansen, Anders Henriksson, Dag Jungenfeldt, Johan Lassing, Filippo Leonini, Thomas Lewin, Yinggang Li, Per Ligander, Fredrik Malmberg, Giovanni Milotta, Sonia Nardin, Stefano Panzera, Andrea Quadrini, Anna Rhodin, Mats Rydström, Daniel Sjöberg, Leonardo Tufaro, Karl-Gunnar Törnkvist and Dan Weinholt.

Acknowledgements

References

1. Radio-frequency arrangements for fixed service systems, Recom-mendation ITU-R F.746-9, Geneva, Switzerland, 2007.

2. Ericsson Annual Report 2010, Stock-holm, Sweden, March 2011. http://www.ericsson.com/thecompany/investors/financial_reports/2010/annual10/sites/default/files/Erics-son_AR_2010_EN.pdf

3. Recommended guidelines for the accommodation and assignment of multimedia wireless systems

(MWS) and point-to-point (p-p) fixed wireless systems in the frequency band 40.5-43.5GHz, ECC recommendation (01)04 (revised), Rottach-Egern, Germany, February 2010.

4. Radio frequency channel arrange-ments for fixed service systems operating in the bands 71-76GHz and 81-86GHz, ECC recommenda-tion (05)07 (revised), Dublin, Ire-land, 2009.

5. J. Hansryd, P-E Eriksson, High-

speed mobile backhaul dem-onstrators, Ericsson Review, 2/2009. http://www.ericsson.com/ericsson/corpinfo/publi-cations/review/2009_02/files/Backhaul.pdf

6. Propagation data and prediction methods required for the design of terrestrial line-of-sight systems, Recommendation ITU-R P.530-13, Geneva, Switzerland, October 1, 2009.

7. Larsson, P., Lattice array receiver

and sender for spatially orthonor-mal MIMO communication, Ve-hicular Technology Conference, 2005. VTC 2005-Spring. 2005 IEEE 61st, vol.1, pp. 192-196, May 30-June 1, 2005.

8. Bohagen, F., Orten, P., Oien, G. E., Construction and capacity analy-sis of high-rank line-of-sight MIMO channels, Wireless Communica-tions and Networking Conference, 2005 IEEE, vol.1, pp. 432-437, March 13-17, 2005.

Jonas Hansryd

joined Ericsson Research in 2008 and is currently a senior researcher in broadband

technologies. His focus is on high- capacity microwave links to meet the data-rate, latency and traffic-volume demands on mobile backhaul created by evolved HSPA and LTE. He has been involved in the development of high-capacity 70/80GHz microwave links, as well as LoS MIMO microwave communication systems. He holds a Ph.D. in electrical engineering from Chalmers University of Technology in Gothenburg, Sweden, and was a visiting researcher at Cornell Universi-ty, Ithaca, US from 2003 to 2004.

Jonas Edstam

joined Ericsson in 1995 and is currently responsible for techno-logy strategies at Product

Line Microwave & Mobile Backhaul, and is also an expert on microwave radio transmission networks. He has many years of experience in this area and has, in various roles, worked with a wide range of topics, from detailed microwave technology and system design to his current focus on the strategic evolution of packet- based mobile backhaul networks and RAN. He holds a Ph.D. in applied solid-state physics from Chalmers Universi-ty of Technology in Gothenburg, Sweden.

FIGURE 6 4x4 SS-DP LoS MIMO outdoor setup

28

E R I C S S O N R E V I E W • 1 2011

Analytics – the truth is in there Understanding consumers’ behavior and their changing needs are key ingredients in good business decision-making. Valuable information in networks is underutilized, and this article describes Ericsson’s approach to data analytics – how to create smart services from data.

information is provided for the process-es used by operators to serve their cus-tomers with maximum efficiency.

It seems that operators are facing the classic combination of a challenge and an opportunity, both arising from the same basic phenomenon. The issue at hand is the surge in the amount of data generated by today’s networks. Operators control an amazing amount of user data, demographic information and mobility patterns. Traffic to, from and among subscribers is yet another valuable source of information – mod-erated, of course, by opt-in, opt-out and other legislative and cultural lim-itations. The challenge: how to turn enormous numbers of bits into useful information? Traditionally, telephony systems have handled millions of call-detail records per day, resulting in the creation of customized management systems. The new traffic era, with dif-ferent kinds of data traveling over air-waves through copper and through fiber, is generating massive amounts of data, stretching system capabilities to the limit. Ericsson’s research into how to manage the data surge and push limits much further is outlined below, as well as how to widen the extent of available information to encompass all data handled by telcos.

From bits to insights Analytics focuses on finding patterns and using complex pattern-matching techniques on new data to make sense of it. The sheer volume of informa-tion in networks makes it impossi-ble for a person to process and absorb the multitude of events and data that is handled by a modern com-munication network. Consequently, the first level of analysis must be computer-processed. Intelligent pro-gramming looks for known patterns,

From a framework perspective, several levels of processing are required before bits become intellect. Figure 1 illustrates the relationships between data (bits), information (data in context), knowledge (interpreted informa-tion) and wisdom (the ability to make the right decisions). Here, Ericsson’s holistic approach to cross-domain analytics, and how it allows many kinds of informa-tion to be interpreted using a common toolset, are discussed in practical terms. This implies that network information can be set into a user and service context, service behavior and acceptance can be related to network condi-tions, and consumer experience measurements can be related to all touch points.

BackgroundInformation enables choice, relevance and accuracy in the decision-making process. Operators have been – and remain – the information gatekeep-ers of the telecom industry. To date, valuable information available in tele-com networks has not been used to its full potential to make smart busi-ness decisions – but this situation is changing rapidly. That change is being driven by the need to shift focus from technology and services to user experience and relevance. As mobile

broadband approaches global ubiquity, operators are spending more time on subscriber retention than acquisition, and this has resulted in them paying greater attention to the following stages of the customer lifecycle:

target – enhance market awareness, proposition targeting and promotion relevance;buy – make the purchasing process smooth and enjoyable;set up – ensure efficient and seamless provisioning of the purchased service;use – guarantee that the service meets requirements for functionality and quality, and that customers are using the services available to them to the greatest possible extent;pay for – provide clear, convenient and easy-to-use payment facilities;support – enhance users’ experience of customer care channels; andgrow – include the flexibility to easily modify contract and service levels as customer needs evolve, or as a result of churn-management activities.

To provide effective management of the customer lifecycle, accurate and timely information is needed. Operators have an excellent opportunity to serve their customers better, but at the same time, they face issues created by the radically increasing amounts of data being gen-erated by modern networks. Ericsson’s focus is to address these issues, by hand-ling the complexity and volume of data from the networks and other sourc-es. In this way, correct and relevant

U L F OL SSON A N D JACO FOU R I E

Finding gems in the data mines


CEP complex event processingENIQ Ericsson Network IQKPI key performance indicatorLTE Long Term Evolution

OSS/BSS Operations and Business Support Systems ToD time of day

BOX B Ericsson Network IQ provides service-aware and network-performance management with statistical trend analysis, by service or subscriber, and problem-management alarms for proac-tive problem indication.

Wisdom

Knowledge

Information

Understanding

Analytics

Consolidation

Data

FIGURE 1 From data to wisdom

29

E R I C S S O N R E V I E W • 1 2011

understanding, implementing change, and benchmarking progress. In other words, reporting is moving from passive formatted reports to full-circle, active decision-making processes in which the periodic report is less important and the process of continuous change is the primary focus; there is a shift from a network-oriented approach – nodes and services – to users and user segments, with a focus on how users experience the services provided by the operator; and increasingly, focus is on finding new segments and innovative offerings that the operator can offer.

Figure 2 shows a high-level layer view of the new analytics architecture, illus-trating how the flows from all infor-mation sources are unified to provide the above capabilities. Information sources exist in all domains of the opera-tor’s network – access, multi-access edge, standard services, custom services, transport, management – and possibly from business-to-business partners (indicated by the bottom row of icons in Figure 2).

Capture and aggregationThis layer aggregates pieces of

encompass both aspects of analytics in a common, efficient and reliable infra-structure.

Sources of informationIn the telco world, traffic-handling nodes are naturally considered as pri-mary sources of information, and they are indeed the source of most data. Increasingly, however, other types and sources of information are being used to create a holistic view of the user, including:

user-related data such as name, demo-graphics (gender and income bracket), account status, subscription history, and customer-reported issues; and service-related data such as frequently used services, average usage duration, from which device and from where.

The Ericsson portfolio includes products such as ENIQ Statistics and ENIQ Events, as well as the Analytics Suite set of tools that build on the comprehensive range of online charging products.

ArchitectureAnalytics today is different from traditional analytics in three ways:

reporting is no longer just about accounting and is instead moving toward

with rules-based detection subsystems integrated with, or at least very close to, the network nodes generating and transferring data. The result of this pro-cess is filtered, aggregated and possibly reformatted data, which then becomes information – data with associated metadata. Data volumes resulting from this level of processing are more man-ageable, but they still tend to be associ-ated with the network elements (or at least domains) where they originated.

The next level of processing is mainly dedicated to detecting instances of dif-ferent types of patterns. Here, the idea is to find patterns relating to network events in context – relating to who the user is, for example, which services they have accessed and what kind of device they are using. This process is, how-ever, essentially about detecting pre-determined patterns. At this point, the number of instances of a certain type of event can be displayed on an opera-tor’s dashboard showing where identi-fied KPIs meet their targets.

This type of analytics can be com-pared to the production side of a mining operation: the whereabouts of the ore is known and its content is known, as is the technique used to extract it. It is not surprising that analytics is sometimes called data mining. However, there is another side to the mining process that has a similar parallel in the data field: namely prospecting.

Prospecting is the process of find-ing previously undiscovered wealth – the skills and tools of the prospector lead to the detection of buried trea-sure. The parallel in data-mining is usually referred to as ad hoc analysis, which involves twisting data around, associating and slicing it in various ways until the analyst finds previously unidentified patterns. Consequently, an interesting duality arises in smart business intelligence originating, on the one hand, from ad hoc analysis, and on the other, from production analytics. Ad hoc analysis requires flexibility, high performance and rapid recalcula-tion of vast amounts of data using intui-tive graphics tools. Production analytics drive the big boards in operation centers and feed business dashboards providing essential business intelligence reports. The challenge – which defines the archi-tecture that needs to be built – is to

Visualization Exposure

Query engine

Capture/aggregation

Stream processing Storage

Business Support Systems

Customer databases and repositories

Network elements 2G/3G/LTE

Analytics Suite

Serviceanalytics

Loyalty andchurn analytics

Campaign andpromotion

management

Customerprofilinganalytics

FIGURE 3 The layered approach to analytics

30

E R I C S S O N R E V I E W • 1 2011

event information into a complete description of an event, enriching the data with relevant user and service information. Much of the existing busi-ness logic (after detecting and filtering out known patterns) can be executed here, significantly reducing the need for processing capacity higher up in the stack and thereby improving the cost-to- performance ratio.

FIGURE 2 Cross-domain analytics architecture

StorageA substantial number of recent devel-opments in analytics have taken place in the area of storage. The main focus of these developments has been on methodologies for handling, stor-ing and manipulating data in order to keep costs down and performance up. Ericsson contributes its telecommuni-cations experience to facilitate linear scaling and maximum flexibility for

both traditional reporting and future analysis activities.

The cost of disk space can make up a substantial proportion of the total spending for an analytics system. Column databases, which radically increase query-execution performance and reduce storage space, are just one example of the technology improve-ments that contribute to reduced cost.

On a different track – sometimes referred to as part of the NoSQL trend – open-source initiatives such as the Hadoop framework and the accom-panying Hadoop distributed file sys-tem are being used to explore other ways of addressing the issue of storage. Typically, these initiatives improve processing performance by enabling massive scale-out possibilities, where processing and storage load is shared across a large number of servers.

A combination of the two approaches seems to be the most beneficial solution: using the Hadoop approach in the lower levels where scale-out properties can be best exploited, and feeding into structured databases for the dashboard, report-generation and exploratory anal-ysis work in the higher layers.

Stream processingReal-time data-processing becomes possible as systems and data move from being costly, isolated and offline to being affordable, linked and online. This, in turn, leads to improved moni-toring of services and faster identifica-tion of pattern changes. This enables immediate and accurate decision- making, which is key for modern telecommunications operators.

In particular, real-time processing can be used to merge streams and use time-based filtering logic to find cas-es where transactions take an unusu-ally long time, for example. Complex event processing (CEP) systems, as they are often known, tend to be high- performance systems with limited intermediate storage, meaning that data needed on a long-term basis must be sent to storage. There is thus a need for communication between stream processing and data storage; the CEP engine usually needs reference infor-mation from the long-term storage area so that it can assess the events that occur.


31

E R I C S S O N R E V I E W • 1 2011

For the storage and stream process-ing layers, Ericsson relies on qualified partners, as these technologies have applications in a wide range of industries and are outside the scope of Ericsson’s core technologies.

Query engineThis level of the system contains the business logic – the rules and process-ing components that turn data sets and filtered events into summaries that can be presented on dashboards and tabulated in reports. The need for flexibility here is high: changing the parameters of a KPI, for example, should be controlled by the enterprise (in this case, the telco). In contrast, operators who want to know what is going on in their networks and with their custom-ers, but who lack a budget for highly qualified analysts, need a vendor with a portfolio of ready-made analytics to pro-vide clear answers to relevant questions.

This is one of the key benefits of the Ericsson Analytics Suite approach: it provides a comprehensive set of pre-integrated solutions to a number of sig-nificant operator issues such as churn prediction and campaign analytics. Figure 3 shows the ever-expanding set of analytics applications supporting operators’ current and future needs.

The query engine enables ad hoc queries to efficiently and intuitively support the search for new users through an analyst’s ability to find undetected patterns. Top performance is essential in this area, as the analyst’s ability to be creative and test ideas is reduced when processing takes several hours to complete.

In this layer, Ericsson’s telco expe-rience is a major differentiator, as domain knowledge is essential to be able to phrase the right ques-tions and to know what informa-tion must be retrieved to answer those questions and how to organize queries for maximum effectiveness.

VisualizationFlawless analytics results are of little use if they cannot be converted into human understanding and correspond-ing action. Effective tools to display bar charts, trend indicators, cluster diagrams or any other kind of visual representation are absolutely vital to

unlock the real potential of knowledge hidden deep in the data.

This is a rapidly evolving area, where high-performance, flexible tools are required. Users should be able to adjust their display to suit the task at hand; futile attempts at accurate guesswork from an engineer’s desk far from the operational realities is unlikely to pro-vide the solutions needed in the flux of day-to-day network operations.

ExposureFinally, analytics results can also be used in other systems, made available as database tables or possibly as online services. User profiles and call-success data are simple examples of informa-tion that can be made available to a customer care workstation in real time.

Aggregated mobility data that is made available for traffic monitoring systems to assess traffic congestion is a somewhat more complex usage scenar-io. In this layer, telco expertise and expe-rience are vital in order to interpret and combine data correctly.

Use cases By using the data in the OSS/BSS domain efficiently, an operator can improve the relevance and efficiency of decision-making in many disciplines, ranging

from product management, marketing and sales to service management and customer care.

The following use case shows how product management and customer care areas can use OSS/BSS data to build customer profiles that provide more efficient and relevant customer care, maximizing customer satisfac-tion, targeting and retention.

Customer profiling is centered on a customer profile card, which is a multi-dimensional view of a user that can be utilized for several different business processes.

The type of information held in a pro-file card includes:

demographic data;behavioral characteristics/segments;device(s) details;broadband profile;social-network score;churn score; andcustomer experience for each product.

The above categories can be further broken down by category-specific attributes. Some examples of attributes for the broadband profile are:

browsing profile – such as preferred browser and average session duration;news profile – including frequency and most frequently used news services; download profile – such as device

FIGURE 4 A typical analytics dashboard

BOX C Profile cards are updated continuously as users con-sume operator services. If a user’s behavior changes due to a change in income, for example, the profile card will change accordingly. Consequently, a profile card supplies the operator with accurate user information, improving the relevance of operator inter-actions with the user. This builds loyalty and gives users a greater sense of intimacy with their service provider.

FIGURE 5 Effect of anti-churn campaign to targeted set of users compared with churn score for a control set of users to whom the campaign was not offered.

32

E R I C S S O N R E V I E W • 1 2011

ume demands envisioned in the long term, an additional level of cutting-edge technology is needed. This is one of the drivers for Ericsson’s significant invest-ment in data-management research.

However, the capacity increase can’t be allowed to interfere with other vital characteristics. Thus, the second chal-lenge is how to reconcile capacity with the flexibility and convenience that the next generation of telco analysts will require. Ericsson’s solution to this is to ensure that logic and presentation can be easily created and modified using flexible and powerful design tools.

Finally, the third challenge will be how to balance the technological pos-sibilities with the ethical and legal lim-itations imposed by business practice and emerging legislation. Monitoring and contributing to EU and other legis-lation is of prime importance, as these laws will help to outline the require-ments for consumer integrity protec-tion functionality such as opt-in mech-anisms. Additionally, Ericsson conducts consumer research to ensure that the systems can be adapted to fit not just formal requirements, but also to meet consumers’ expectations on how, when and for what purposes their data should be used.

ConclusionEricsson already has the technology and products in place to solve a wide range of telco analytics challenges, combining the capabilities of network, service and user analytics. The aim is for the com-pany to evolve these capabilities con-tinuously, in dialog with customers to provide the best, most cost-effective and most adaptable telco analytics solution on the market.

updates, content including music or movies, volume and ToD; and shopping pattern – such as shopping services visited.

By examining the values held in a sub-scriber’s profile card, customer care rep-resentatives can, for example, gain an understanding of how users experience the services they subscribe to, and eval-uate how valuable a specific customer is to the operator and whether a user is at risk of churning due to poor service experience.

Using profile cards enables targeted campaigns and advertising based on the profile card scores. For example, a specific user profile might indicate that the user is a strong influencer in their social network and is interested in tennis. Based on this information, a spe-cial advertisement that includes tickets to a local tennis event can be sent to that user. This user is more likely to spread this information further within their social network than a user with a low social-network score.

Operators can use the churn score to target potential churners with specific retention campaigns to try to reduce the risk of these users churning. Analytics Suite enables operators to identify potential churners based on:

experience – the user’s satisfaction

level based on their experience of the operator’s services;social-network score – taking into consideration the churn score of influential people in the user’s network (indicating whether they have a high churn score or have churned recently);account status, including prepaid balance information, postpaid bill data and pattern tracking information;refill and bill-payment data highlighting changes in patterns, indicating an increased churn risk; andchanges in usage patterns.

This data is used to set churn scores that can be used by product management, marketing and sales to effectively target retention activities and campaigns to specific users. Churn scores are updated continuously to provide simple bench-marks for monitoring the effectiveness of user-retention activities. By simply comparing the before and after values, operators can estimate the effectiveness of each activity. ChallengesThe first – and possibly most obvious – challenge is raw performance. Today’s technology provides enough capacity to handle both current data volumes and volumes predicted for the imme-diate future. However, to meet the vol-


33

E R I C S S O N R E V I E W • 1 2011

Jaco Fourie

began his career in data communication, specializing in the X-Series Standards. He joined MTN

South Africa in 1994 with focus on value-added services and charging. In 1999, he joined Ericsson, working with product management for charging systems. He is now senior expert and head of Business Development and Strategy with Business Unit Multimedia, Solution Area Revenue Management, where he focuses on the future evolution of the Ericsson BSS portfolio to suit the demands the Networked Society will put on next-generation BSS.

Ulf Olsson

has a background in software architecture for distributed military Command and Control

systems. He joined Ericsson in 1996, working mainly with architecture issues concerning packet-based systems, such as packet PDC, GPRS, UMTS and the CDMA2000 packet core network. He is now senior expert with Business Unit Multimedia, Systems Management group. He focuses on data management, business support systems, IMS and developer-oriented issues. He holds an M.Sc. in engineer-ing physics from the Royal Institute of Technology (KTH), Stockholm.

34

E R I C S S O N R E V I E W • 1 2011

Operator-provided visual communication Visual-communication technology has existed since the 1960s in various forms, including ISDN video-conferencing, internet-based desktop solutions, and immersive telepresence rooms. The technology is well established among enterprises yet, for several reasons, it has not been adopted on a mass-market scale.

The inclusion of all criteria – AV quality, reasonable cost of providing adequate AV quality, interoperability and reliability – in a system, calls for standardization that forms the basis of affordable, non-proprietary, high-quality visual-communication services. This article outlines the proof of concept (PoC) deployment at Ericsson of an evolved multimedia telephony service based on the established 3GPP IMS standard. The deployment shows that IMS can support a visual-communication system that meets the requirements for audio and visual quality, interoperability, reliabil-ity and cost, thus facilitating the entry of visual-communication services into the mass market.

IMS platformIMS is the standard framework for IP-based multimedia services and was consequently chosen as the platform for the visual-communication PoC deployment. IMS supports some basic functions that are not yet support-ed by other proprietary systems. Such functions include a global address-ing scheme, QoS control of access and transport, and standardized inter-working principles. GSMA has adopt-ed an IMS-based VoLTE solution for voice and messaging, and has recent-ly begun work on a complementary IMS profile for visual-communication services over LTE mobile access. IMS is ideally positioned to support the deploy-ment of visual-communication services on mobile broadband devices such as smartphones and tablets – a factor

Prerequisites for the mass-market adoption of visual communication are good audio and video (AV) quality and optimum user experi-ence at an appropriate cost. Trials have shown that users begin to appreciate visual-communication services once AV quality reaches a certain level. Although solutions that provide an acceptable level of quality are available, the equip-ment remains prohibitively expensive, preventing the large-scale adoption of services.

The processing capacity required for HD video and high-quality audio is already present in many consumer products, including mobile phones. The cost of peripherals, such as displays, cameras

and microphones, has declined signif-icantly in a very short time. The conve-nience of components that are cheaper and commercially available has led to the development of high-quality visual- communication end points.

Interoperability and reliability are additional obstacles that have prevented the technology from being adopted for mainstream use. In order for visual- communication technology to be taken up by the mass market, users need to be able to connect to other users via completely reliable connections, irrespective of the network and plat-form they are using. User demands are beginning to drive development, and include desired functionality such as seamless switching to and from visual communication from other media, including voice.

Quality to get you talking


3GPP 3rd Generation Partnership ProjectAV audio and videoAVC advanced video-codingCCAS conference control application server DSL digital subscriber lineEVS Enhanced Voice Services GSMA Global System for Mobile Communications AssociationGUI graphical user interfaceHD high-definition HSPA High-Speed Packet AccessIMS IP Multimedia SystemIP Internet ProtocolIPTV IP Television

ISDN integrated services digital networkITU International Telecommunication UnionLTE Long Term EvolutionMCU multipoint conference unitMRFP media resource function processorNNI Network to Network InterfacePoC proof of conceptQoS Quality of ServiceSD standard-definition SIP Session Initiation ProtocolSVC scalable video codecVoLTE voice-over-LTEUNI User to Network Interface

BO BU R M A N, A N DE R S E R I K SSON, A N DR E A S BE RGQV I ST, K L AUS SCH N E I DE R A N D H Å K A N DJ U PH A M M A R

Complexity

Bit rate

Quality

FIGURE 1 H.264 video quality

35

E R I C S S O N R E V I E W • 1 2011

consideration to ensure a positive experience for the user.

Ericsson’s PoC IMS deployment uses the H.264 advanced video-coding (AVC) video compression standard. This codec is widely used in areas as diverse as digital television, Blu-ray, YouTube and 3GPP mobile communication sys-tems. AVC supports video quality rang-ing from thumbnails at rates of a few kbps to cinema-screen-sized full HD at several hundred Mbps. Ericsson Research contributed to the standard-ization of AVC, and recently optimized an internal version to support real-time simultaneous encoding and decoding of full-motion HD-quality video on a PC.

As with most other video-coding stan-dards, AVC generates a video stream with a variable bit rate. However, the output rate can be controlled through an encoder-internal feedback loop, the algorithm for which is not standardized. Consequently, Ericsson’s PoC implemen-tation includes several different types of rate-control algorithm, ranging from IPTV to low-delay conversational. The AVC standard allows considerable free-dom to set encoding parameters, mak-ing it possible to trade processing

poor-quality audio, frequency respons-es or echo control. The end points used in the PoC implementation include state-of-the-art speech enhancement functions such as full-stereo echo- cancelling, noise reduction and time-scaling jitter-buffers.

VideoAdding media to complement voice on an IMS-based system is a fairly straight-forward process. The technology’s flexibility allows a voice call between two video-capable terminals to be transformed into a voice-and-video call by simply adding video media to the call. Similarly, a voice-only terminal can reject video media in an incoming multi media call, and the call remains voice-only.

Video communication involves many of the same challenges as voice, such as firewall traversal, jitter and end-to-end latency. Video, however, is more sensitive to packet loss due to high codec compression. There is a trade-off between bandwidth and quality; determining the minimum acceptable bandwidth with respect to quality for various usages of video is an important

that is instrumental in securing mass- market adoption of this technology.

Included in IMS are standardized definitions for interworking princi-ples on the UNI and NNI interfaces, and IMS also provides a well-defined integration point, which can be lever-aged to achieve interconnectivity with existing proprietary systems.

Proof-of-concept implementation AudioSource-signal bandwidth has a greater impact than anything else on audio quality. The human voice spans a fre-quency range of at least 50Hz-12,000Hz, and human hearing spans a similar or even larger range. Full reproduction of the human voice requires a sample rate of 32kHz or 48kHz.

Spatial listening in a telecommuni-cation environment can be emulated using stereo capture and rendering. For audio-only clients, the additional spatial information that comes with stereo rendering improves the listener’s ability to determine who is speaking. For video clients, it further enhances the experience because there is a tight-er correspondence between the video image and the audio stream. In other words, the voice of the person speaking appears to come from the direction of the screen displaying the video of that person. Stereo further improves the per-ceived audio quality in rooms with less-than-ideal acoustics.

Ericsson’s PoC IMS deployment uses stereo-audio based on G.719 encoding (an ITU standard audio codec). This recently standardized codec delivers a 48kHz sample rate at affordable bit rates and provides high-quality audio to fixed-access network users. Ongoing standardization activities for a new 3GPP codec, known as Enhanced Voice Services (EVS), will provide similar levels of audio quality at bit rates suit-able for mobile-access networks.

Audio quality in real-time commu-nication is affected by several other factors, including the acoustic envi-ronment, the electroacoustic hard-ware, speech enhancement, and trans-mission losses. End points should be designed to take each of these fac-tors into consideration to ensure good audio quality. Subsequent network processing cannot compensate for

36

E R I C S S O N R E V I E W • 1 2011

requirements for video quality at a given bit rate, as illustrated in Figure 1.

The video codec parameters for the PoC implementation were selected using ITU Standard P.910, Subjective Video Quality Assessment Methods for Multimedia Applications. These assessment methods are used to eval-uate multimedia applications to select algorithms and assess quality during an audiovisual connection.

Tests were conducted in native and up-scaled resolutions, and video qual-ity was assessed as a function of reso-lution, bit rate and codec parameters. Professional high-quality video record-ings from typical conferencing situa-tions were used as source material for the tests.

UsabilityAn essential part of the PoC system design process were usability studies. The conclusion of these studies was that high-quality media must be com-plemented with ease of use if visual communication is to gain widespread popularity.

The influence of quality and ease of use on the PoC trials is exemplified in the way pre-booked multi-party video calls are handled. At startup, all client-related booking information is retrieved from the central booking system. Booking information is continuously updated as long as the client is registered. Booked meetings are dis-played in the GUI, and the user can join the desired meeting by selecting it from the list displayed.

The booking function for meetings has been integrated with e-mail and calendar systems through the develop-ment of a plug-in for Microsoft Outlook. This plug-in allows users to add video to a regular meeting simply by click-ing a button in the calendar appli-cation. Other aspects of the PoC user interface, such as the address book and call-handling procedures, were designed after consideration of the results of the usability studies.

ArchitectureAs a consequence of Ericsson’s PoC trial results, two additional functions were added to IMS to extend its capabilities: a conference control application server (CCAS) and a media resource function


P-CSCF

BGF

MRFPConferenceAS

HSS SLFIBCF

MGCF

I/SCSCF

IMS/SIP/H.323

interworking

Fixedbroadband

access Ut/XCAP SIP/ISUP

IP/TDMSIP

RTP

Mobilebroadband

access

Mobilenarrowband

access

PLMN/PSTN interworking

MGW

AS application serverBGF Border Gateway FunctionCSCF Call Session Control FunctionHSS Home Subsriber ServerI/S CSCF interrogating/serving CSCFIBCF Interconnection Border Control FunctionIMS IP Multimedia SubsystemIP Internet ProtocolISUP ISDN user partMGCF media gateway control functionMGw media gateway

MRFP media resource function processorP-CSCF proxy-CSCFPLMN public land mobile networkPSTN public switched telephone networkRTP Real-time Transport ProtocolSIP Session Initiation ProtocolSLF Subscriber Location FunctionTDM time-division multiplexingUt lossless video codecXCAP XML Configuration Access Protocol

FIGURE 2 IMS architecture

Transcoding/mixing

Video resolutionHighMediumLow

MRFP

HD client

HD client

HD client Single stream client

FIGURE 3 MRFP video handling

37

E R I C S S O N R E V I E W • 1 2011

processor (MRFP). These functions provide a simple mechanism to man-age clients participating in a visual- communication session.

A visual-communication session can be scheduled as a multi-party (confer-ence) call that each client connects to, or an ad hoc multi-party call initiated by one client. The CCAS instructs the MRFP, which is the IMS term for a multipoint conference unit (MCU), to create a voice and video bridge for the clients, while the MRFP ensures that all participating clients have audio and visual contact.

For multi-party calls, there are several ways of distributing media to clients, and each method has its advan-tages and disadvantages. In the distrib-uted conference method, each client sends all relevant media to all other clients. All communication is sent directly between peers and requires no specific media support in the net-work, minimizing delays. The main disadvantage of this approach is that the total bandwidth consumed by the call rises exponentially with each addition-al client. Furthermore, unless a com-mon codec with a shared configuration is used, each client must encode the media individually to each peer, plac-ing a significant computational load on the sender.

An alternative method – centralized conference – lets each client make a point-to-point call to an MRFP in the network. Bandwidth demands increase linearly with each additional client, and provided that the MRFP can support a large set of codecs and configurations, each sender can use their preferred media encoding. The disadvantages of this approach are:

participants have less freedom to compose and present media;the complexity of the MRFP can become significant;media is often delayed while being processed; and media processing can cause quality degradation due to the transcoding introduced when composing, resulting in a mosaic video image of the participants.

The IMS PoC implementation is based on a combination of centralized and distributed methods, exploiting the advantages of each method to the great-est possible extent. For audio, the MRFP

terminates the media plane and mixes the audio signals from all active partici-pants, in the same way as it would dur-ing a regular centralized conference. For video, however, the media is switched to the furthest possible extent without re-encoding it, as shown in Figure 3.

To facilitate video switching without the need for transcoding in the MRFP, a number of common formats have been adopted. HD-capable clients send three video streams – HD, SD and thumbnail – to the MRFP. Standard-definition clients send two streams: SD and thumbnail. In this way, clients with reduced pro-cessing capacity or limited-access band-width can just send thumbnails. The MRFP relays the appropriate streams to the other clients based on speech activity or manual user choice.

In the PoC system, a client receives one high-resolution stream – HD or SD, depending on the capabilities of the cli-ent – and several thumbnail streams. The client can then present full-screen video of the active participant with thumbnail overlays of the non-active participants. The system permits any combination of video images to be nego-tiated between the client and the MRFP.

Because each client sends multiple video streams, the screen layout of each participant can be tailored, while at the

same time reducing the transmission delay and avoiding any loss of quali-ty due to transcoding in the MRFP. A similar solution could be implement-ed using the H.264 scalable video codec (SVC); however, with the conditions at hand – video streams with significant differences in resolution and a require-ment for efficient PC implementation of the codec – the AVC solution is a more advantageous one.

The PoC implementation includes two end points: a client intended for fixed installation in conference rooms and a personal client for desktop or laptop use. Both types are implement-ed as soft clients on standard PCs and use affordable off-the-shelf consumer electronics for AV delivery.

InteroperabilityAn essential part of building a visual- communication service for mass- market deployment is interoperability. Mobile telephony would not have achieved the market success it enjoys today if it had been based on non- compatible, proprietary protocols. In contrast, the visual-communication market continues to be dominated by numerous solutions that are mutually incompatible, creating an acute need for a single standard.

FIGURE 4 Typical video-communication display where the active speakers are shown in full screen and other speakers in thumbnail images

38

E R I C S S O N R E V I E W • 1 2011

Several major players in the mar-ket have already initiated standardiza-tion activities. The use of IMS, however, solves this problem. As an open stan-dard, IMS is available to all vendors, and unlike many proprietary solutions, it can be implemented on systems from mobile devices to telepresence rooms, and over different network architec-tures such as DSL, HSPA and LTE.

IMS provides standardized user-to-network and network-to-network interfaces that allow operators to create networks that support multi-media features beyond messaging. IMS meets user demands for seam-less switching to and from visual com-munication from other media such as voice in a generic and interoperable way. Ericsson’s IMS deployment uses a stan-dardized SIP interface, which is becom-ing increasingly popular for enterprise video systems.

Reliability – the operator’s roleOperators have a key role to play in visual communication. Because real-time visual communication requires audio and video synchronization, it is vital that this service is provided on a high-quality core network with low packet loss and low latency. For mass-market breakthrough of visual com-munication reliability and trust are key issues; and operators are uniquely positioned to offer this service. By providing reliable and secure visual- communication services, operators can reduce subscriber churn from internet-based competition while driving IP traffic into networks where it can be controlled and monetized.

SummaryAn evolved multimedia telephony service using IMS has already been deployed, using commercially avail-able, affordable PC and audiovisual equipment. The PoC deployment start-ed in 2010, and there are now more than 40 multimedia conference rooms installed in Ericsson facilities world-wide. Feedback indicates a high level of user satisfaction, and numerous requests for additional installations have been received.

The success of the project shows that an IMS-based system meets each of the criteria – sound quality, video quality,

interoperability and easy switching of media – that users require from a visual-communication system, with operators supplying the remaining piece of the puzzle: reliability.

The IMS standard is ideally suited to support the expansion of the visual- communication market from the enter-prise to the consumer sector. Similarly, visual communication can become the mass-market service that IMS requires to become widely adopted by operators – and that will result in sat-isfied users who can finally enjoy the service quality that will make visual communication mainstream.

The next stepEricsson’s PoC has demonstrated the possibilities offered by an IMS-based visual-communication solution. It is clear that the potential for an operator-provided, high-quality mass-market service for video meetings is great.

This conclusion has led to the ongo-ing work of bringing ideas from the PoC to the market. With close coopera-tion between two of Ericsson’s business units, BU Networks and BU Multimedia, a commercial solution is currently being defined.

Bo Burman

is currently working as a senior specialist in video telephony at Ericsson Research Multimedia

Technologies. Since joining Ericsson in 1996, he has worked with video com-pression and various aspects of mobile video communication in multiple research and business unit projects, including standardization in 3GPP, ITU-T, IETF and MPEG. He holds an M.Sc. in computer technology and engineering with a focus on telematics from the Institute of Technology at Linköping University, Sweden.

Andreas Bergkvist

is strategic product manager in Product Line IMS with a focus on short- and long-term strategy

development for securing a competi-tive Ericsson offering in the video com-munication area. He is also solution manager for the VisualCom program. He joined Ericsson in 2005 on the R&D Global Graduate Program and has since had various roles, primarily in the development organization for the Ericsson Session Border Gateway, in-cluding project management, system management and technical product management. He holds an M.Sc. in engineering physics from Uppsala University, Sweden.

Anders Eriksson

is an expert in audio media processing at Ericsson Research. Since joining Ericsson in 1994 he

has been active in research, standard-ization, and implementation of speech signal processing aspects of the voice service. He holds an M.Sc. in applied physics and electrical engineering from Linköping University, Sweden, and a Ph.D. in automatic control from Uppsala University, Sweden.


39

E R I C S S O N R E V I E W • 1 2011

Håkan Djuphammar

is vice president of Architecture & Portfolio at Group Function Techno-logy & Portfolio Manage-

ment, and is responsible for overall architecture and portfolio issues for the Ericsson Group. He holds an M.Sc. in telecommunications from Imperial College, London, UK and a B.Sc. in electrical engineering from Chalmers University, Gothenburg, Sweden. Pre-viously, he was vice president, Product Line Management within Ericsson’s Business Unit CDMA Mobile Systems. He has held various positions at Erics-son since 1991, including some in the areas of 3G product management and PDC system management. Before join-ing Ericsson, he was co-founder and president of Abstract Electronics AB.

Klaus Schneider

joined Ericsson in 1991 and is currently working as program director for the VisualCom program with a

focus on product planning and indus-try relations. He has held various posi-tions at Ericsson R&D in development, project management, and line man-agement, including head of a product development unit with responsibility for introduction, deployment and sup-port of Ericsson’s Core and IMS prod-uct range. He holds an M.Sc. in mathe-matics from the Technical University (RWTH) of Aachen, Germany.

Re:view

25 years ago In 1986, Ericsson Review

published an article on the frequency planning of digital radio-relay networks. Similar to the article in this issue on microwave capacity evolu-tion, the older article men-tions that the task of the network-planning engineer is to select radio frequencies and antenna types in such a way that the influence of interfering signals is within the margins of the planning objectives for the overall performance of the radio circuit.

50 years ago In 1961, an article was

published about extending telephone plants with regard to value of subscribers time. While this article essentially covered the cost and dimen-sioning aspects of plant expansion, it was based on figures measuring the incon-venience to subscribers as as result of congested net-works. The article mentioned subscribers’ demands in terms of accessibility and in-telligibility of conversations and offered a method of deter-mining the number of switches needed. Interestingly, the method conluded that the value of subscribers’ time had the greatest impact on dimensioning.

75 years ago In 1936, Ericsson Review not-

ed that 60 years had passed since Lars Magnus began the activities that would lead to the creation of the Ericsson Group. An article depicting the evolu-tion of Ericsson’s telephone in-struments between 1878 and 1935 showed how tastes and raw-material technologies had shaped the telephone.

Ericsson Review, issue number 2, 1986.



Telefonaktiebolaget LM EricssonSE-164 83 Stockholm, SwedenPhone: + 46 10 719 0000Fax: +46 8 522 915 99

ISSN 0014-0171Edita Västra Aros, Västerås

Documents

The communications technology journal since 1924 1/2011 · PDF fileThe communications technology journal since 1924. 1/2011. Heterogeneous networks increasing cellular capacity . 4