Static Lightpath Routing in IP over WDM Networks Universität Würzburg/ITG Fachgruppe 5.. Workshop IP Netzmanagement, IP Netzplanung, und Optimierung Würzburg, Germany - July 4, 00 Dominic Schupke Munich University of Technology Institute of Communication Networks 8090 Munich, Germany Tel.: +49 89 89-35 Fax: +49 89 89-635 Email: Schupke@lkn.ei.tum.de
Overview Motivation and Introduction Node and Network Models Solution Approach Finding the Virtual Topology Wavelength Assignment Graph Coloring Case Study Results Conclusions
Exponential growth of IP backbones
Motivation IP fiber backbones - usage of WDM: Optical crossconnects (s) interconnect IP routers by lightpaths No intermediate multiplexing layers (e.g., SDH, ATM) less administration, ) fixed capacities per lightpath Aim: determine efficient (static) configuration of the lightpaths for IP router demands, i.e. find an optimal virtual topology for IP and a routing and wavelength assignment (RWA)
Optical Crossconnect Models N WC Source Traffic selectable Sink Traffic variable F WC Source Traffic selectable Sink Traffic variable P WC Source Traffic selectable Sink Traffic variable N N N N N N from other nodes M j i M to other nodes from other nodes M W variable M Converters fixed to other nodes from other nodes M C M C to other nodes 3 C selectable Converterpool variable {N,F,P}-WC = {Non,Full,Partial}-Wavelength Converting
Main Characteristics The Entire Problem Flow Formulation Route Formulation IP Layer WDM Layer + Unknown virtual topology Metric assignment + Partial conversion Long computation time + Path constraints Path pre selection + Path constraints Wavelength constraint Large problem ) approach: consider subproblems
Network Model IP over N-WC WDM IP Router IP Router 0 3 IP Router s are interconnected by fiber pairs (wavelength constraint) are connected to at most one IP router (source and sink for least one lightpath) Transponders or colored interfaces No wavelength converters present (wavelength continuity constraint) Shortest path based IP routing protocols (e.g., OSPF) Equal IP routing metrics (e.g., all )
Solution Approach Routers, Demands Virtual Topology Routing for Lightpaths s, Fibers Wavelength Assignment Graph Coloring Whole problem becomes too complex (let alone RWA) ) subproblems solved each by heuristic algorithms: Objective: minimize laser sources (significant cost factor) Virtual topology: minimize virtual links (and thus laser sources) Routing of lightpath demands: shortest path (keeps fiber usage and lightpath lengths low) Assignment of wavelengths: graph coloring problem (even minimization of number of wavelengths)
Finding the Virtual Topology Bandwidth Demands G = G V D Calculate used Link-Bandwidth using Shortest Path Routing Aim: find virtual topology with minimum number of links such that bandwidth demand is carried Find (next) Link L with min. used Bandwidth no By deleting Link L, yes Constraints violated? G = G \ L G = G V V V V Further constraints: link capacity maximum number of hops (keeps IP packet delay low) maximum amount of transit traffic through an IP router (avoids overwhelming of routers and reduces IP packet delay)
Wavelength Assignment and Graph Coloring Network example: Corresponding path graph: IP Router 3 0 3 0 0 3 3 0 IP Router 0 IP Router 0 3 0
Graph Coloring Algorithm Aim: assign colors to the path graph nodes such that adjacent nodes do not have the same color the number of used colors is minimized Algorithm: Degree of Saturation (DSATUR) heuristic Degree of saturation: the number of colors that are not allowed for a node (adjacent nodes have already obtained these colors) A= all uncolored nodes with max. degree of saturation Take a node of Awith the max. degree Assign the first available color to this node Mark the neighbor nodes with the color label and delete the adjacent edges
Case Study NSF network with 4 nodes (node: IP router plus ) and links (fiber pairs) Bit rate capacity of a wavelength: 0 Gb/s Traffic matrix: scaled such that entry with maximum value equals 0 Gb/s (total network load: 0 Gb/s) Palo Alto Seattle 3 San Diego 4 Salt Lake City Boulder 5 6 Houston 7 Lincoln Pittsburgh 8 Urbana Ann Arbour 0 9 Atlanta 4 Ithaca College Park Princetown 3
Results: Dependence of Allowed IP Hops Number of Virtual Links 00 89 75 50 3 5 9 0 3 Number of IP Hops 4 Number of Wavelengths 6 4 4 0 8 6 6 4 5 4 0 3 4 Number of IP Hops ) -3 hops: already reasonable results
Results: Dependence of Allowed Transit Traffic Traffic matrix: full mesh demand with demand-pairs of Gb/s Number of Virtual Links 00 9 75 69 57 50 5 5 7 4 7 Number of Wavelengths 6 5 4 3 0 8 6 6 4 5 7 7 0 0 5 0 5 0 5 30 35 40 45 50 55 60 65 70 75 80 Maximum IP Forwarding Traffic in Gb/s heuristic nature of the virtual topology algorithm Restrictions can be used to obtain result alternatives. ) 0 0 5 0 5 0 5 30 35 40 45 50 55 60 65 70 75 80 Maximum IP Forwarding Traffic in Gb/s
Conclusions Static lightpath routing in IP over WDM networks: Finding a virtual topology for the IP layer Routing and wavelength assignment on the WDM layer The results of the heuristic algorithm approach show: The simple heuristic for virtual topology produced already reasonable results DSATUR algorithm for the graph coloring performed well Due to heuristic approach: short execution time Further information: http://www.transinet.de/