WWW workload generators are used to evaluate web server performance,
and thus have a large impact on what performance optimizations are
applied to servers. However, current benchmarks ignore a crucial
component: how these servers perform in the environment in which
they are intended to be used, namely the wide-area Internet.
This paper shows how WAN conditions can affect WWW server performance.
We examine these effects using an experimental testbed which emulates
WAN characteristics in a live setting, by introducing factors such
as delay and packet loss in a controlled and reproducible fashion.
We study how these factors interact with the host TCP implementation
and what influence they have on web server performance. We demonstrate
that when more realistic wide-area conditions are introduced, servers
exhibit very different performance properties and scaling behaviors,
which are not exposed by existing benchmarks running on LANs. We show
that observed throughputs can give misleading information about server
performance, and thus find that maximum throughput, or capacity, is a
more useful metric. We find that packet losses can reduce server capacity
by as much as 50 percent and increase response time as seen by the client.
We show that using TCP SACK can reduce client response time,
without reducing server capacity.
|