22
NetBoxIT: an Open Testbed for Heterogeneous Networks Study Giorgio Giorgio Calarco Calarco , , Maurizio Maurizio Casoni Casoni Department of Engineering “Enzo Ferrari” University of Modena and Reggio Emilia Italy

NetBoxIT: an Open Testbed for Heterogeneous Networks Studymcasoni/pres_IWCMC12.pdf · NetBoxIT: an Open Testbed for Heterogeneous Networks Study Giorgio Calarco, Maurizio Casoni Department

Embed Size (px)

Citation preview

NetBoxIT: an Open Testbed forHeterogeneous Networks Study

Giorgio Giorgio CalarcoCalarco, , Maurizio Maurizio CasoniCasoni

Department of Engineering “Enzo Ferrari”University of Modena and Reggio Emilia

Italy

Maurizio Casoni IEEE IWCMC – Limassol (Cyprus), 30 August, 20121

Outline

• Introduction: Emergency Networks

• The FP7 Large Scale IP E-SPONDERVisionNetwork Architecture

• NetBoxIT: a modular, flexible, open framework for the study of heterogeneous networks

• Experimental results from the test-bed

• Conclusions

Maurizio Casoni IEEE IWCMC – Limassol (Cyprus), 30 August, 20122

FP7 Large Scale IP E-SPONDER

Title:A holistic approach towards the development of the first responder of the future

Objective: SEC-2009.4.2.1: First Responder of the future

Duration of the project: 48 months (started July 1st 2010)

Maurizio Casoni IEEE IWCMC – Limassol (Cyprus), 30 August, 20123

“system of systems”architecture

Personal Area Personal Area NetworkNetwork

First responderpersonal network;interconnects

terminals, sensors,…

IncidentIncident Area Area NetworkNetwork

Temporary network;  Set up by MEOC;Dati between users

and MEOC;

JurisdictionJurisdiction Area Area NetworkNetwork

Main Emergency net;Fixed infrastructures;IAN traffic managmnt;

ExtendedExtended Area Area NetworkNetwork

Nation wide opticalfibre networkBackbone for First 

responders, IAN, JAN

Public Safety System ArchitecturePublic Safety System Architecture

Maurizio Casoni IEEE IWCMC – Limassol (Cyprus), 30 August, 20124

General Architecture

FRs normally act either in remotely located areas with limited or disrupted communication infrastructures.

They need to exchange information with the Mobile Emergency Operation Centre (MEOC) and with the remote Emergency Operation Centre (EOC), to enable cooperation at all levels with the target to minimize the uncertainty typical of crisis events.

Maurizio Casoni IEEE IWCMC – Limassol (Cyprus), 30 August, 20125

The E-SPONDER Network Architecture

Main backhaul link via satellite

Extended area network (EAN), acts as a backbone for JANs

Jurisdiction area network (JAN), fixed infrastructures, eventually used as backup backhaul links

Incident area network (IAN), mesh network serving on-field FRs

Personal area network (PAN), wireless sensors collecting environmental information

Maurizio Casoni IEEE IWCMC – Limassol (Cyprus), 30 August, 20126

Mobile Emergency Operation Centre

A vehicular communication infrastructure to supportcommunications among FRs, other MEOCs, and the EOC

Possible technologies:

• DVB-RCS NG (main backhaul to the EOC)• WiMAX (inter-MEOC mesh)• WiFi or femtocells (FRs)• Bluetooth (sensors)

• 3G, 2.5G, TETRA (backup backhaul)802.11 802.16 802.15

SAT TETRA 3G

MEOC

Maurizio Casoni IEEE IWCMC – Limassol (Cyprus), 30 August, 20127

Possible Approaches for Performance Evaluation

Analysis: traffic theory• Ideal but difficult for complex scenarios

SimulationPros: flexible, repeatable, cheapCons: huge amount of time and computing resources for complex

scenarios, not so accurate and useful for implementation

Physical test-bedPros: realisticCons: expensive to deploy, not suitable for “a priori” evaluations

Hardware emulatorsPros: based on “ad hoc” devices: realistic and suitable for

heterogeneous network studiesCons: expensive and usually not open to researchcommunity

Maurizio Casoni IEEE IWCMC – Limassol (Cyprus), 30 August, 20128

A modular & programmable framework

“Ideal” tool:• as flexible as software simulator• as modular as hardware emulator• as realistic as physical test-bed

Our proposal:hybrid approach, by using network emulators with virtualization techniques

We could virtualize each part of the emergency network using severalsimulators/ emulators and interconnect them through software entities

Maurizio Casoni IEEE IWCMC – Limassol (Cyprus), 30 August, 20129

Benefits of a hybrid approach

Maurizio Casoni IEEE IWCMC – Limassol (Cyprus), 30 August, 201210

NetBoxIT

NetBoxIT thought as trade-off:• based on commodity hw• “off-the-shelf” sw• accurate modeling of networks• support of real-time emulation• modular: complex scenarios throughcombination of simple elements

Novelties:1. Containers used to virtualize emulators

and run several of them in parallel• Assign private CPU-corse to each

emulator2. Ethernet is used to bind emulators

together and to external entities

a netbox represents a network portioncreated by running a network emulatorwithin a dedicated virtual machine

Maurizio Casoni IEEE IWCMC – Limassol (Cyprus), 30 August, 201211

Which software tools ?The wireless networks simulation: NS-3, one of the fastest simulators around, can also act as an emulator, allows real-world testbed integration

Virtual Machines: traditional virtualization techniques allow to run several concurrent operating systems; examples: VmWare, VirtualBox, etc. This technique usually requires lots of computing resources

Containers: light virtualization technique, allow to virtualizea whole operating system as well as a single application -Sun Containers, LXC

Why virtualization ? Using Containers/VMs we can assign to each NS3 instance a controlled amount of resources (CPU, memory, network interface, etc), as if NS3 was running on a different PC

Maurizio Casoni IEEE IWCMC – Limassol (Cyprus), 30 August, 201212

NetBoxIT is based on

• HardwareDual Xeon Quad Core E5530 2.4GHz 8 physical CPU-cores (two CPUs with 4 cores)PCI-Express I/O bus4 Intel Gigabit NICs

• Software: Linux 2.6 kernelLinux Containers (LXC 0.7.4) to hostNS-3.9 emulators

Hardware

Virtual Machine Host or Containers

802.11s emulator

SAT emulator

802.16 emulator

3G emulator

802.15 emulator

TETRA emulator

Maurizio Casoni IEEE IWCMC – Limassol (Cyprus), 30 August, 201213

Heterogeneous network: the reference network scenario

Run of several netboxes (networks)In parallel on a single platform

Components:

External PC: voice flow from FRWi-Fi netbox: IEEE 802.11a FR-FRCWiMax netbox: FRC-MEOC linkSat netbox: return-forward channelsExternal PC: EOC equipment

Return/Forward2/10 Mbit/s

Ideal: 254 ms latency

-76000 KmDVB-RCS-NGMEOC-EOC

Return/ForwardChannel:18/18 Mbit/s

FRIIS30 dBm2.5 Km802.16 TDD10 MHz @5 GHz

FRC-MEOC

54 Mbit/sFRIIS16 dBm80 mIEEE 802a FR-FRC

TheoricalGross Capacity

ChannelModel

TX Power

DistanceTechnologyNetworkSegment

Maurizio Casoni IEEE IWCMC – Limassol (Cyprus), 30 August, 201214

Maximum IP goodput (return channel)

0

1000

2000

3000

4000

5000

6000

7000

8000

100 500 1000 2000 4000 8000

IP Throughput (Kbit/s)

Input rate (Kbit/s)

Wi‐Fi

Wimax

SAT

UDP flows500 bytes IP packets

Performance:as expected(for WiMax as well!)

NetBoxIT provides realistic results

Maurizio Casoni IEEE IWCMC – Limassol (Cyprus), 30 August, 201215

Traffic profiles and settings from FR/FRC to EOC

-754676Total

256 bytes305272CBR Data

256 bytes401388CBR Video

74 bytes4816CBR VoIP

Packet sizeIP rate(Kbit/s)

Source rate(Kbit/s)

Settings for QoS on the WiMax segment

• VoIP managed with rtPS• VIP with UGS (highest priority)• Data as Best Effort

Maurizio Casoni IEEE IWCMC – Limassol (Cyprus), 30 August, 201216

VIP vs. VoIP single flow

254.4

261.5

265.0266.7

276.5 274.3

240.00

245.00

250.00

255.00

260.00

265.00

270.00

275.00

280.00

VoIP (rtPS) VIP (UGS)

Late

ncy

(ms)

VoIP vs. VIP Latency

Min

Average

Max

0.5 0.4

3.94.4

6.0 5.7

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

VoIP (rtPS) VIP (UGS)

Jitte

r (m

s)

VoIP vs. VIP Jitter

Min

Average

Max

Maurizio Casoni IEEE IWCMC – Limassol (Cyprus), 30 August, 201217

VIP vs. VoIP: 2 parallel flows

254.7

261.5

267.7 266.9

286.2

274.2

235.00240.00245.00250.00255.00260.00265.00270.00275.00280.00285.00290.00

VoIP (rtPS) VIP (UGS)

Late

ncy

(ms)

VoIP vs. VIP Latency

Min

Average

Max

0.5 0.4

8.3

4.4

10.8

5.6

0.00

2.00

4.00

6.00

8.00

10.00

12.00

VoIP (rtPS) VIP (UGS)

Jitte

r (m

s)

VoIP vs. VIP Jitter

Min

Average

Max

Maurizio Casoni IEEE IWCMC – Limassol (Cyprus), 30 August, 201218

VIP vs. VoIP with 8 data flows

272.4254.9 261.4

313.5

269.9 267.6

410.1

310.9 301.4

0.00

50.00

100.00

150.00

200.00

250.00

300.00

350.00

400.00

450.00

Data (BE) VoIP (rtPS) VIP (UGS)

Late

ncy

(ms)

Bulk Data, VoIP, VIP Latency

Min

Average

Max

3.11.0 0.5

9.4 8.3

5.2

32.5

17.7

11.9

0.00

5.00

10.00

15.00

20.00

25.00

30.00

35.00

Data (BE) VoIP (rtPS) VIP (UGS)

Jitte

r (m

s)

Bulk Data, VoIP, VIP Jitter

Min

Average

Max

Maurizio Casoni IEEE IWCMC – Limassol (Cyprus), 30 August, 201219

VIP vs. VoIP with 8 data flows

292.3 258.4 263.8

1001.8

341.0

269.5

2280.3

536.8

304.2

0.00

500.00

1000.00

1500.00

2000.00

2500.00

Data (BE) VoIP (rtPS) VIP (UGS)

Late

ncy

(ms)

Data, VoIP, VIP Latency

Min

Average

Max

3.4 1.2 0.5

53.5

17.1

5.4

110.6

31.0

13.7

0.00

20.00

40.00

60.00

80.00

100.00

120.00

Data (BE) VoIP (rtPS) VIP (UGS)

Jitte

r (m

s)

Data, VoIP, VIP Jitter

Min

Average

Max

FRC-MEOC distance increased to 7.5 km => maximum capacity = 775 Kbit/s1 VIP + 2 VoIP + 8 data = 802 Kbit /s => WiMax link saturated

VIP flow preserved, VoIP flows degraded, data flows very delayed, as expected

Maurizio Casoni IEEE IWCMC – Limassol (Cyprus), 30 August, 201220

Conclusions

• NetBoxIT: test-bed for the study of heterogeneous networksLinux containersNS-3 as simulator/emulator

• It has been shown to be• Modular: several netboxes can be run in parallel with no penalities• Interoperable and real time: binding with real networks, devices and apps• Flexible: any kind of network can be emulated with NS-3 and

virtualized• Open: use of Linux kernel 2.6, LXC and NS-3• Scalable: by assigning a CPU-core to every virtualized network

• Future worksinvestigating even more complex networksanalyzing other tools (i.e. netmap) for performance improvementsemployment of software router, e.g. Click!,in the test-bed

Maurizio Casoni IEEE IWCMC – Limassol (Cyprus), 30 August, 201221

THANK YOU FOR YOUR ATTENTION

[email protected]@ieee.org

http://www.dii.unimo.it/casoni

… suggestions are very very welcome

Department of Engineering “Enzo Ferrari”University of Modena and Reggio Emilia