Multi-domain routing techniques in ASON networks

Abstract

A pan-European ASON network has been simulated under dynamic traffic to compare the different multi-domain routing procedures allowed by the ASON Recommendations. Compared features are: network performance (blocking probability) and scalability (amount of required routing-information). Introduction According to ITU-T ASON Recommendations [1-3] different routing strategies may be adopted in a multidomain network environment to distribut routing information between the domains While several papers deal with Multi-Domain Routing (MDR) issues in a multi-layer network environment [46], only few works have been specifically dedicated to inter-domain routing in ASON. Our study, developed in the framework of the IST European Project MUPBED (Multi-Partner European Test Beds for Research Network) (http://www.ist-mupbed.eu), aims at carrying out a scalability-versus-performance comparison of the different ASON inter-domain routing approaches under dynamic traffic. Routing performance has been measured by computing the blocking probability of the connections as a function of the traffic load, while the scalability of each MDR strategy has been evaluated by estimating the amount of routing information it requires. ASON routing models An ASON network is partitioned into Routing Areas (RAs). Typically, a RA cover an administrative domain. The synchronized “ensemble” of nodes of an RA behaves as a single abstract entity called Routing Performer (RP), maintaining detailed intra-area routing information and a synthetic and summarized view of the topology of the other RAs. For inter-area (inter-domain) routing, ASON Recommendations [1,2] describe the kind of routing information exchanged by the RPs (link-state (=with attributes: e.g. weight, occupancy) or reachability (=purely topological) information) and how path computation is distributed between the RPs across the network (source-based or step-by-step (=area-by-area) path selection). Elaborating on the standard (and assuming each connection having a pair of User Network Interfaces (UNIs) as end-points) we have defined the following set of routing procedures. SBS-R1: each RA propagates only reachability information regarding UNIs it contains; step-by-step path selection is adopted. SBS-R2: similar to SBS-R1, but each RP knows the shortest-path RA sequence to a destination UNI. SRC-L2: each RP distributes linkstate information on “tunnels” (i.e. forwarding adjacencies) for transit inter-area connections; path selection is source-based. SRC-L3: similar to SRCL2, but each RP propagates link-state information also for tunnels from its border to its internal UNIs, i.e. the source has metric information also inside the destination RA up to the destination UNI. SRC-LA: every RP has a complete, detailed, link-state-based representation of the whole network; path selection is source-based. As benchmark, we have also considered a single-domain (-area) scenario (SDMLA) in which the whole network is managed as a single RA: the only difference with SRC-LA is that in SRC-LA intra-area connections (having UNI endpoints both in a given RA) are bounded to be routed within the area itself (can only use internal links of the domain). Case-study network and MDR-method complexityevaluation by the novel parameter TTE Simulations have been carried out using the network topology represented in Fig. 1: it is composed by a simplified version of the GEANT2 European research network interconnected to simplified versions of the 5 National Research and Education Networks (NRENs) participating to the MUPBED project. Each one of these 6 subnets is assumed to be a domain (an ASON RA). Each NREN is interconnected to GEANT2 by external Node Network Interfaces (eNNIs) and to other two NRENs via eNNIs located on cross-border links (these links do not exist in