THE EFFECTIVENESS OF AFFINITY-BASED SCHEDULING IN MULTIPROCESSOR NETWORK PROTOCOL PROCESSING (EXTENDED VERSION)

Techniques for avoiding the high memory overheads found on many modern shared-memory multiprocessors are of increasing importance in the dev elopment of high-performance multiprocessor protocol implementations. One such technique is processor-cache affinity scheduling, which can s ignificantly lower packet latency and substantially increase protocol processing throughput [30]. In this paper, we evaluate several aspects of the effectiveness of affinity-based scheduling in multiprocessor n etwork protocol processing, under packet-level and connectionlevel par allelization approaches. Specifically, we evaluate the performance of the scheduling technique 1) when a large number of streams are concurr ently supported, 2) when processing includes copying of uncached packe t data, 3) as applied to send-side protocol processing, and 4) in the presence of stream burstiness and source locality, two well-known prop erties of network traffic. We find that affinity-based scheduling perf orms well under these conditions, emphasizing its robustness and gener al effectiveness in multiprocessor network processing, In addition, we explore a technique which improves the caching behavior and available packet-level concurrency under connection-level parallelism, and find performance improves dramatically.