Document

Test results
We ran WebBench’s default ecommerce API test suite, which generates both secure and non-secure static and
dynamic HTTP 1.0 GET requests. While running the ecommerce suite, the clients must negotiate to a secure
Web server port using the Secure Socket Layer (SSL) protocol. WebBench incrementally increases the number of
clients making the HTTP 1.0 GET requests to the Web server. As the workload increases the number of clients,
the Web server’s processor utilization also increases, until the processor in the Web server is saturated with work.
Each workload point with a fixed number of clients is a WebBench “mix”. The ecommerce API test suite begins
with a mix that involves one client, then a mix with four clients, and then increases the number of clients by four
with each mix to a total of 60 clients. A standard WebBench run thus involves 16 mixes.
A WebBench run reports the total requests per second a server can perform and the total throughput, in bytes per
second, that the server delivered. WebBench reports these results for each mix. A graph of these results yields a
performance curve with a peak at some number of clients.
Figure 2 shows the WebBench peak results in requests per second and in throughput (bytes per second) of the
test servers.
Each result is the median of three runs.
Server
Requests
per Second
Throughput
(Bytes per
Second)
Intel Server Board SE7520AF2 with Dual-Core Intel Xeon processor 2.80GHz 28,837.8 116,951,440.9
Intel Server Board S5000PAL with Dual-Core Intel Xeon processor 5160 40,679.7 164,785,622.9
Figure 2: Median requests per second and throughput results for the two test servers running WebBench. Higher numbers
are better.
2
Principled Technologies, Inc.: Intel Server Board S5000PAL with Intel
Xeon processor 5160 WebBench 5.0 performance report