Research On Performance Evaluating Metrics Of Multi-port Routers

Routers and switches are key relaying devices, which route IP messages or switch data-link-layer frames in sub-networks. To verify conformity to relevant protocol implementations in a router or switch, it needs to pass conformance tests. While for evaluation of their performance, quantifiable parameters pertinent to network QoS, such as throughput, transit delay, jitters, and data loss, etc., have to be tested in an environment reflecting the real operation conditions. Therefore, apart from appropriate test methods and test systems controlling test process, there is a need for test metrics, which define what types and probability distribution of traffic should be injected into inputs of a router/switch, how to control or coordinate traffic for individual inputs, and what pattern should be used in directing traffic to outputs in testing. This is exactly the topic dealt in this MSc thesis. The work presented in this dissertation is backed up by the research work at SC-Netcom Lab concerning test methodology and test system for multi-port routers and switches.IETF has defined a set Request For Comments (RFC) related to performance testing for routers and switches, such as RFC1242, 1944, 2544, 2285 and 2889, among them, RFC2544 and 2889 are the most updated. However, the metrics defined in these standards are largely set for a situation, where traffic from different inputs is switched to all outputs evenly. In other words, they did not cover a more critical situation, where traffic is switched to all outputs unevenly. RFC2285 and 2889 only made suggestions for partially meshed traffic situation where multi-traffic is instantly directed to the same output port. Moreover, it is not well defined how to test some parameters such as jitter. This has spurred the research work aiming for generic metrics for testing routers and switches.For the very reason, the author has simulated evenly and unevenly switching pattern with data streams of Constant Bit Rate (CBR), Variable Bit Rate (VBR), and Available Bit Rate. The simulation results show that the main performance parameters are quite different with the two switching patterns, and to(?) extent prove the author's presumption. This has led to an expansion to existing IETF performance testing metrics. As a contribution, the work presented in this dissertation provides a basis for test case definition of performance testing in the laboratory testing activities and a foundation for new proposal to IETF concealing test metrics.