Plan if 2000

  • Upload
    kajfahm

  • View
    218

  • Download
    0

Embed Size (px)

Citation preview

  • 8/11/2019 Plan if 2000

    1/5

    Planif 2000, a tool for planning SDH networks

    Carlos S da Costa*, Joo Henrique Duarte Preto Paulo**, Pedro Azevedo***

    *IT/ISCTE, Av. Rovisco Pais n 1, 1049-001 Lisboa, Portugal, Email: [email protected] , Tel: +351-21 841 84 88

    **MOTOROLA Portugal, Av. Jos Gomes Ferreira, Ed. Atlas I, 9, sala 22, 1495 Algs Portugal,Email: [email protected], Tel:+351 96 296 8256.

    ***SIEMENS Portugal, S.A., IC-OCN N-Network Management Rua Irmos Siemens, 1, 2720-093 Amadora, Portugal,

    Email: [email protected], Tel: +351 21 424 2509

    Abstract

    This paper describes the software tool "Planif 2000" that

    implements a set of different algorithms to help a

    Synchronous Digital Hierarchy (SDH) network planer in

    choosing the final solution or a small set of solutions for

    Synchronous Digital Hierarchy networks. A description ofthose algorithms is presented, and at the end, a relevant case

    study is presented.

    I. INTRODUCTION

    SDH networks replaced old Plesiochronous Digital

    Hierarchy (PDH) infrastructure of established operators and

    is the solution for new operators. The SDH technology has

    important advantages and possibilities for expansion and

    combination with new technologies compared with the

    previous technology, the PDH. SDH is a standard, supports

    rates as high as 40Gbit/s, needs less equipment to extract

    each tributary and each individual channel, exhibits high

    configuration flexibility of services and lines. SDH allowsseveral new types of protection, which increase the system

    availability, SDH technology supports a network

    management entity (TMN). SDH allows the equipment

    compatibility between different vendors.

    When it is necessary to plan the SDH infrastructure to

    cover a metropolitan region with several dozens of nodes

    there is a multiplicity of solutions with different cost. An

    exhaustive analysis of all the alternatives is time consuming

    and inefficient so some heuristics are used to reduce the

    number of solutions to a manageable number and even so the

    number of solutions is large because many parameters can be

    changed, e.g. number of nodes per ring, maximum capacity

    of each link, etc.

    The operator demands from the planner an efficient

    solution, one that minimises the cost assuring a good quality

    of service and protection. The inputs for such planning

    process are the cartographic coordinates (or instead the

    relative positions of the nodes), the traffic between nodes

    and, optionally, some other restrictions. The output consists

    of a set of SDH rings, a set of point-to-point connections a

    list of equipment and the global cost.

    "Planif 2000" is an auxiliary tool that implements a set of

    different algorithms to help the planer in choosing the final

    solution or a small set of solutions.

    The "Planif 2000" converges towards a solution for a SDHcoverage of a region consisting of a central SDH ring of high

    capacity that interconnects several secondary SDH rings, and

    some point-to-point connections. "Planif 2000" assures that at

    least two links end at each node so no node is isolated in case

    of a single link failure.

    "Planif 2000" allows the user to create/edit the network

    through the user interface, and to visualise the net after each

    algorithm allowing the planner to tune some parameters, likenumber of nodes per link, viewing each net structure by

    itself, run each algorithm to check results or run all

    algorithms in one go. In some steps, the user can choose from

    more than one algorithm that can be applied usually carrying

    to different results. At the end, the list of necessary SDH

    equipment and its cost is presented to the user for the global

    solution or for each structure.

    II. SDH NETWORK PLANNING

    When planning an SDH network the mechanism that

    increases the reliability demands from the planner special

    care due to the different protection mechanisms that he can

    choose and the equipment requirements that each of them

    demand.

    The complexity of SDH networks due to its flexibility

    versus PDH has the following items:

    Grouping. While PDH cares about individual

    multiplexing sections with the goal of reducing the

    number of multiplexing units, SDH trails can be

    optimised to convey the tributary units in containers

    through chosen paths.

    Equipment. The PDH equipment has a fixed attribution

    between tributaries ports and output ports. SDH

    flexibility is achieved by a switching matrix that allows

    reconfiguration between tributaries and outputs asneeded.

    Switching. PDH cross connection is carried manually

    while SDH equipment performs cross connection

    automatically.

    The SDH planning process has to take in account:

    The coexistence between high rate SDH and low rates

    PDH nets.

    The continuous increase of demand of

    telecommunications services.

    The market dynamic which makes difficult long-term

    forecasts.

    The rapid technological evolution that shortens theequipment life time.

  • 8/11/2019 Plan if 2000

    2/5

    The SDH is pushed by two streams of evolution, the

    telecommunication stream and the computer stream.

    The existence of several telecommunication operators

    instead of a single one.

    The high reliability required from the equipment and

    from the networks.

    The application of SDH is carried at sub-network levels:

    Central network, between transit central office

    characterised by high capacity, high physical diversity,

    with very special recover requirements and efficient

    usage of the available bandwidth.

    Regional network, between primary central offices

    highly connected, with medium capacity, limited

    physical diversity and specific recover and flexibility

    requirements.

    Access network, to connect access nodes with star or

    tree topologies characterised by low capacity, low

    physical diversity and specific flexibility requirements.

    The planing process can be decomposed in several steps to

    simplify the study:

    The topological analysis, where the network connectivity

    and flexibility are evaluated

    The traffic analysis, with the analysis of the distribution

    and general load of the network traffic.

    The routing, where the most proper path for each

    element in the traffic matrix is calculated. In this process

    several politics can be used: reducing the number of

    jumps, choosing the shortest path, chosen two disjoint

    paths, traffic balancing. The ring is the most common

    structure.

    The architecture and management, The protection looks for traffic share, Multiplexing

    Section Protection (MSP), path protection and specific

    protections for ring configurations.

    The grouping this process goal is to concentrate traffic

    and to do so it uses the layer structure so it is possible to

    concentrate between end-nodes, sub-connection and

    connection

    The equipment planning is carried out after the calculation

    of the necessary network functions taking in account

    equipment limitations and reducing the number of extracting

    steps in high capacity nodes.

    III. THE PLANIF2000 ALGORITHMS

    Planif 2000 final goal is to study several ways of planning

    to achieve a final solution for a telecommunications network

    where criteria like lower cost, better capacity, less utilisation

    of network elements and the grater or less number of rings

    and point to point connections are used. This is carried by

    Plannif2000 using several algorithms with different

    complexity.

    Topoinit Algorithm is used to initialise the traffic matrix.

    The first approach is to consider a fully connected network.

    This network has too many connections some of them not

    cost effective. A simplification of this fully connected

    network is needed to keep the computation time of the trafficrouting algorithm within acceptable values. In this procedure,

    connections with no traffic are eliminated as well as long

    distance connections usually with low traffic. Although there

    is the danger of elimination a better solution that could be

    find if an exhaustive approach was used. The limit of having

    always a node with two connections was imposed to assure

    redundancy connectivity. This connection matrixsimplification can be achieved using one of two algorithms:

    The Crossing Algorithm analyses exhaustively all node

    connections and considers the relative position and the

    survivability of the network when removing

    connections. This way, connections between nodes

    working in parallel (or redundant) are eliminated [1].

    The Average distance between connections algorithm

    performs an exhaustive analysis of all network

    connections. Connections are eliminated if the distance

    exceeds a value corresponding to the average of the

    distance of the connections of two nodes between the

    two nodes in analyses. This value is weighted by a

    factor enabling different solutions with variable

    number of connections[1].Selection of Hubs . As a secondary analysis, it is possible to

    identify nodes with higher traffic characteristics or higher

    connectivity and the algorithm detects those nodes by an

    exhaustive search. Those nodes with special properties will

    usually concentrate traffic and will be used as gateway nodes

    that switch traffic to a, to be designed, central ring. Two

    algorithms were implemented:

    One chooses hubs based on the number of connections

    between nodes after the initialisation. The number of nodes

    that are classified as hubs are in an interval of percentage

    defined by the user.

    The other algorithm works based on the quantity of traffic

    that a node switches and is dependent on the initial traffic

    matrix and on the total number of hubs chosen[1].Selection of point to point exclusive connections . The goal

    of this algorithm is to remove from next optimisation steps

    connections that could increase unnecessarily the future

    rings capacity. The user defines from which hierarchy or

    from how many connections of one hierarchy the connection

    should be considered exclusive1point to point connection.

    Ring generation . Rings are generated using the traffic

    affinity algorithm. This algorithm starts by looking for the

    node with more traffic terminated and that has not started anyring before. Then it chooses other nodes by traffic affinity

    until it reaches the maximum number of nodes in a ring or the

    defined percentage of traffic load of a ring.Simulated annealing is used to organise each ring generated

    in the previous procedure. This algorithm finds the shortest

    cyclic path between graph nodes.Central ring selection , It is crucial in a network that every

    element of the traffic matrix is routed, hence the necessity of

    having an element of network support where all rings and

    1Exclusive is used because this traffic will be transported apart from

    the rest and will not impact further the planning process.

  • 8/11/2019 Plan if 2000

    3/5

    connections are linked together to assure the previous

    requirement. Three methods were implemented to select a

    central ring:

    To select nodes with higher possibility of connecting

    with more rings.

    The selection based on nodes with have more rings incommon.

    The selection based on nodes with more affinity of

    traffic terminated.

    The user has to select which method to use, and the number

    of nodes of the central ring, or the percentage of occupancy

    of the central ring.

    Secondary rings selection . The goal of this stage is to find

    the secondary rings to transport traffic between nodes not

    belonging to the central ring. The rings must always have two

    nodes that belong to the central ring due to redundancy and to

    assure connectivity between all nodes. Two different

    algorithms were implemented to select the secondary rings.

    Secondary Rings-1, this algorithm uses the same principle

    used in the generation of rings starting by the node with more

    traffic terminated and choosing the others by traffic affinity.

    This process forces that two nodes from the central ring are

    chosen so there is redundancy to prevent faults. The

    algorithm stops when all nodes belong to a ring

    Secondary Rings-2, this algorithm starts by the node with

    more traffic not belonging to the central ring. All the other

    ring nodes are selected based on traffic affinity with the

    restriction of having two nodes belonging to the central ring,

    as previously.Routing and consolidation. This algorithm optimises the

    network routing assuming that the net is only constituted bypoint to point connections, and the advantages of having

    rings are not considered. This algorithm works over the

    network connections resulting from the Topoinit algorithm

    and the hubs already selected. It reroutes the 2Mbit/s traffic

    through the higher capacity connections to be able to

    eliminate some connections. The final goal is a network with

    concentration of 2Mbit/s channels on STM-N channels with

    the maximisation of N [1].

    The Final solution selection algorithm is the combination of

    several algorithms: the use of clusters for ring selection,

    network routing with rings, ADM structures and cost

    calculation. These steps are performed in sequence: The first step is initialisation and is necessary to obtain

    the physical connections between nodes (uses the

    Topoinit algorithm).

    The second step is to find exclusive connection and the

    traffic of those connections is temporarily removed

    from the traffic matrix.

    The third step consists of using the concept of cluster

    to group nodes with higher traffic affinity.

    The fourth step consists of finding a central ring.

    The fifth step consists of generating secondary rings

    with a number of nodes chosen by the user. The

    algorithm used here is the one explained before but

    considering the clusters.

    The sixth step consists of selecting the secondary nodes

    that best correspond to the clusters.

    The seventh step consists of selecting point-to-point

    connections to connect isolated nodes, as the previous

    step to assure complete network connectivity. For

    reliability reasons, the user can select redundancy forthese connections.

    The eighth step consists of filling the nodes ADMs

    with the traffic exchanged either with the PDH network

    or with the other nodes. Depending on the type of

    routing (the Dijkstra shortest path algorithm or

    minimising the number of jumps between rings) there

    are different values to ADM filling.

    The ninth step consists of including the traffic of the

    point to point exclusive connections found in step two.

    The tenth step is the calculation of the cost of the final

    solution based in a cost model using ADMs with

    STM1,4,16,64 capacity.

    IV. CASE STUDY

    The case study consists of a metropolitan network with 21

    nodes with the traffic represented in a 2Mbit/s circuit

    requirement matrix. The network data was from a

    Telecommunications Operator (Portugal Telecom) and

    represents the traffic transportation needed for an urban area

    with mixed services (data and voice).

    The case study network showed a heavily loaded

    connection between two central nodes where the exclusive

    connection step was applied.

    In addition, some of the nodes situated at the border of the

    network area were chosen to carry traffic into the trafficmatrix, in order to handle external traffic requirements.

    The traffic matrix used also considers the future growth of

    the network so the 2Mbit/s circuit values are future

    requirements.

    When choosing the best results from each of the set of

    results given by the algorithms the planer tried to follow

    these rules:

    Never chose network solutions where a lot of the

    network traffic is carried by an only few

    connections (choke points).

    Never use rings with its nodes geographically very

    distant or opposite on the network as the case study

    is a metropolitan network. Do not overuse node to node connections as they do

    not take advantage of the SDH ring benefits

    (protection and management).

    Good distribution of the equipment over the entire

    network.

    Even using these restrictions, the planner will get several

    possible network solutions and he must choose the best fit for

    his network.

    Figure 1 shows the result of applying the selection of Hubs

    algorithm. In the figure 1a), hubs were selected between all

    nodes if they have 10 to 20% less traffic than the node with

    more connections. This results in a relative low number ofhubs; four hubs marked in figure 1 a). The second case the

  • 8/11/2019 Plan if 2000

    4/5

    percentage is between 30 and 40%. This result in increasing

    the number of hubs. Seven hubs were found as can be seen in

    figure 1b).

    a)

    b)

    Figure 1 Hub selection a)10 to 20%b) 30 to 40%

    The application of the Selection of point to point exclusive

    connections to the metropolitan network results in the

    connections marked in figure 2 a) where a percentage of

    filling of 20% and at least one STM16 were the parameters

    selected.

    a)

    b)

    Figure 2 Selection of point to point exclusive connections a)

    with 20% of filling STM16 b) with 80% of filling with STM4

    Figure 2b) is the result of the same algorithm when a

    percentage of 80% of filling and STM 4 circuits are selected

    is. The jump to a higher hierarchy is carried out when there isthree circuits from the previous hierarchy.

    The result of applying the ring generation algorithm with

    crescent number of nodes per ring is the convergence to an

    optimal ring; this can be seen comparing figure 3.

    a)

    b)

    Figure 3 Ring generation with a percentage of filling of 80%.

    a) with four nodes per ring b) with nine nodes per ring

    Figure 4 is the result of secondary ring generation where

    can be seen that all secondary nodes are connected by two

    nodes to the central ring. In figure 4 a) secondary rings were

    generated without following the topology limitation and in

    figure 4 b) they were generated following the topology

    limitation.

    a)

    b)

    Figure 4 Secondary ring generation with eight nodes per ring

    a) without following the topology limitation. b) following thetopology limitation.

  • 8/11/2019 Plan if 2000

    5/5

    Comparing the initial topology with ten hubs, figure 5 and

    the result of applying the consolidation algorithm, figure 6

    can be seen a concentration of traffic in the connections

    between hubs, solid lines.

    Figure 5 Initial topology with ten hubs.

    Figure 6 Result of the consolidation algorithm with weight

    parameter equal to twenty.

    a)

    b)

    Figure 7 Final solution for the case study a) standard solution

    b) the best solution tested.

    The final solution presented in figure 7 were obtained using

    the following parameters: percentage of hierarchy occupancy:

    80%; number of exclusive point-to-point connection: 1;

    number of clusters: 3; max. number of nodes per cluster: 3;

    min. number of nodes per cluster: 2; number of nodes in

    central ring 5; max. number of nodes in secondary rings: 6;

    min. number of nodes in secondary rings: 5; with topological

    optimisation and USHR/SNCP protection.

    The following results provided by Planif are used to chose

    among some of the solutions:

    Number of STM-X structures (where X is 1..64 and

    structures are rings or connections) (see table below)STM

    64STM

    16STM

    4STM

    1ADM

    64ADM

    16ADM

    4ADM

    1

    a) 1 4 3 2 5 20 6 4

    b) 1 4 2 2 5 20 4 4

    Number of ADMs of STM-X hierarchy used.

    Percentage of rings versus point to point

    connections. a) 0,66 b) 0,8

    Central ring traffic vs. overall central ring capacity.

    a) 0,64 b) 0,64

    Cost of the solution, rings and connections (vendor

    dependent)

    To find the best optimised solution, Planifs default valueswere used to generate a network solution that was used to

    compare all the other solutions too.

    The main conclusion of this program when applied to the

    Portugal Telecom metropolitan network is that the best

    relation of capacity vs. cost is achieved when the secondary

    rings have between 5 and 6 nodes. This number of nodes

    assures that the hierarchy used is not too high, and there is no

    ring overlapping.

    IV. FINAL REMARKS

    The design of a reliable, cost effective, upgradeable

    telecommunication network is not an easy task for the

    planner and has become more difficult due to the newrequirements and available technologies. With this tool, the

    planner can easily try several solutions changing several

    parameters, methods and algorithms.

    Planif 2000 was developed to be a tool to aid the planner in

    his task of choosing a network solution but will not by itself

    generate the best solution. The network planner must study

    the output it provides and decide on the best relation between

    cost implementation and management of the final solution.

    V. REFERENCES

    [1]Joo Paulo, Pedro Azevedo, Planeamento de Redes

    SDH, IST 1999.

    [2] Sexton, M. e Reid, A., Broadband Networking, Artech

    House, 1997.

    [3] Sexton, M., SONET Interworking, Artech House, 1996.

    [4] Geyer M., SDH Network Applications and related

    Planning Issues, SIEMENS, NETPLAN 10, September 1996.

    [5] Cardwell, R.H., Wu, T.H., Wasem, O.J., Survivable

    SONET Networks Design Methodology, IEEE Journal on

    selected areas in communications Vol. 12, n.1, January

    1994.

    [6] Sexton, M., Soto, O.G., Tardini, C., Wulf-Mathies, C.,

    Planificacin y gestin de redes SDH, Comunicaciones

    Elctricas Alcatel 4 trimestre 1993.