Summary
In this paper, we have presented evidence supporting the effectiveness
of affinity-based scheduling in multiprocessor network
protocol processing. We have 1) evaluated the performance of
affinity-based scheduling while varying the number of admitted
streams, 2) considered performance for protocol processing which
includes the overhead of copying uncached packet data, 3) evaluated
affinity-based scheduling of parallelized send-side UDP/IP/-
FDDI processing, and 4) examined the impact of stream burstiness
and source locality as captured by the Packet-Train source model.
We find that affinity-based scheduling performs well under these
conditions, emphasizing its robustness and general effectiveness in
multiprocessor protocol processing. In addition, we have explored
stream-scaled CLP, which scales the number of independent stacks
with the number of admitted streams, and found the performance
improvement to be dramatic.