Parallel Processing of Protocols
Many workstations of today consists of a multiprocessor system with a shared memory and a small number of processors. We present a “processor-per-message” partitioning approach to do parallel processing of protocols. We have used the approach to implement a shared memory multiprocessor implementation of the x-kernel protocols and execution environment from University of Arizona. The two major performance bottlenecks in shared memory multiprocessor execution of protocols are lock contention and contention for shared memory. Measurement results on the implementation and simulation results of the locking effects are presented. The measured speed-up for the parallel implementation compared to the sequential one is more than 12 times for UDP and 3 times for TCP.
KeywordsParallel Processing Shared Memory Parallel Implementation Multiprocessor System Shared State
Unable to display preview. Download preview PDF.
- Clark, D., Jacobson, V., Romkey, J., and Salwen, H., “An Analysis of TCP Processing Overhead”, IEEE Communications Magazine, June 1989.Google Scholar
- Zitterbart, M., “High-Speed Protocol Implementations based on a Multiprocessor-Architecture”, Proc. 1st IFIP Workshop on Protocols for High-Speed Networks, Zürich 1989.Google Scholar
- Björkman, M. and Gunningberg, P., “Locking Effects in Multiprocessor Implementation of Protocols”, ACM SIGCOMM 93Google Scholar
- M. Goldberg, G. Neufeld, M. Ito, “A Parallel Approach to OSI Connection-Oriented Protocols”, Proc. 3rd IFIP Workshop on Protocols for High-Speed Networks, Stockholm 1992.Google Scholar