A scalable simulation platform for switching and scheduling

Simulation has become a significant way for performance evaluation in information technology research. However, current simulators such as NS-II, OPNET and so on all have few modules of in switching or scheduling and have somewhat limitations in inheritability and scalability. By employing system level design method and object oriented technology, we design and implement a simulation platform called SPES for evaluation of switching and scheduling in which typical switching fabrics and corresponding scheduling policies are integrated. Inheritability and scalability are obtained by separating traffic sources, switching fabrics and scheduling policies as modules entirely, and each module is provided to users by basic class libraries so it is easy for users to explore or append new modules. Simulation results of typical scheduling algorithms exhibit SPESpsilas excellent performances. Furthermore, we make some research on Clos switching fabric which will be the next step in scaling current crossbar switches into many hundreds or few thousands of ports. By establishing the model in graph theory, we get the condition for memory-space-memory Clos to emulate OQ with a broad class of packet scheduling algorithms, including WFQ and strict priorities. Perhaps most interestingly, this result holds for all traffic arrival patterns. After it, we propose a triple round-robin based dispatching (TRRD) algorithm for MSM Clos switching fabric which is very simple and efficient due to its three arbiters. Simulation based on SPES proves TRRDpsilas good performance: high throughput respective of traffic model and lower time delay.