FIRST TRY

CS290F Fall 2006 - UCSB Computer Science - Thorsten von Eicken

Jump to: navigation, search
[edit]

HTTPERF Benchmark (Before Optimization)

      • The Graph below is misleading since during this simulation, httperf's "sessions" parameter was set to 10, while the simulation above was set to 50.

HTTPERF Benchmark Test on Path1

Image:before_path1_graph_01.jpg

Image:before_path1_graph_02.jpg

HTTPERF Benchmark Test on Path2

Image:before_path2_graph_01.jpg

Image:before_path2_graph_02.jpg

HTTPERF Benchmark Test on Path3

Image:before_path3_graph_01.jpg

Image:before_path3_graph_02.jpg

Analysis

In the experiment above, the parameters for httperf was set to be

httperf --hog --server=domu-12-31-33-00-03-da.usma1.compute.amazonaws.com --wsesslog=10,0,<script file> --session-cookie --print-reply --rate $rate

where $rate is monotonically incremented from 0.1 to 5.

After observing the data above, it appears that the maximum request rate a single server can handle is around 22 req/s, and expected response time is about 250 ms when the server gets overloaded. In the future development, We hope to increase the limit of the current maximum request rate and decrease the average response time as well as eliminate such a radical jump in the response time over the request rate, as we witnessed in the Path 1 simulation.

Retrieved from "http://www.voneicken.com/courses/ucsb-cs290f-fa06/index.php/FIRST_TRY"
Personal tools