Forum OpenACS Q&A: Server Performance Benchmarks with Apache Benchmark and Traffic Jamme

1: Server Performance Benchmarks with Apache Benchmark and Traffic Jamme

Posted by James Thornton on 04/29/01 08:58 PM

Compare my Apache Benchmark result of 6.73 requests per second to 745 requests per second documented at https://openacs.org/forums/message-view?message_id=15300.

I am using the nsunix module to host several sites on one IP, and here are my system specs....

PIII 500 MHz
320 MB memory
RAID1 IDE (8ms access time)
RedHat 6.2
AOLserver 3.2
OpenACS 3.x

Here are my results from Apache Benchmark using Apache's default page...

NOTE: I changed .html to .shtml so that it won't be processed as an ADP page (obviously AOLserver is not processing it as a SSI page).

fusion /usr/local/apache/bin 47% ./ab -n 700 -c 50
http://jamesthornton.com/apache.shtml
This is ApacheBench, Version 1.3c <$Revision: 1.41 $> apache-1.3
Copyright (c) 1996 Adam Twiss, Zeus Technology Ltd,
http://www.zeustech.net/
Copyright (c) 1998-1999 The Apache Group, http://www.apache.org/

Server Software:       
AOLserver/3.2                                      
Server Hostname:        jamesthornton.com
Server Port:            80

Document Path:          /apache.shtml
Document Length:        2504 bytes

Concurrency Level:      50
Time taken for tests:   104.051 seconds
Complete requests:      700
Failed requests:        0
Total transferred:      2027900 bytes
HTML transferred:       1752800 bytes
Requests per second:    6.73
Transfer rate:          19.49 kb/s received

Connnection Times (ms)
              min   avg   max
Connect:        0     1    20
Processing:  1756  7187 19640
Total:       1756  7188 19660

With Traffic Jamme, I would get 4 pages / sec (adp pages with database queries).

Here are results from a website on the same server that is not using OpenACS but is being virtually hosted via the nsunix module (246.57 requests per second is better, but still not 700)...

fusion /usr/local/apache/bin 35% ./ab -n 700 -c 50
http://www.wendisanford.com:80/apache.html
This is ApacheBench, Version 1.3c <$Revision: 1.41 $> apache-1.3
Copyright (c) 1996 Adam Twiss, Zeus Technology Ltd,
http://www.zeustech.net/
Copyright (c) 1998-1999 The Apache Group, http://www.apache.org/

Server Software:       
AOLserver/3.2                                      
Server Hostname:        www.wendisanford.com
Server Port:            80

Document Path:          /apache.html
Document Length:        2504 bytes

Concurrency Level:      50
Time taken for tests:   2.839 seconds
Complete requests:      700
Failed requests:        0
Total transferred:      1904810 bytes
HTML transferred:       1757180 bytes
Requests per second:    246.57
Transfer rate:          670.94 kb/s received

Connnection Times (ms)
              min   avg   max
Connect:        0    41    74
Processing:    66   133   169
Total:         66   174   243

2: Response to Server Performance Benchmarks with Apache Benchmark and Traffic Jamme (response to 1)

Posted by Jerry Asher on 04/30/01 12:01 AM

Hi James,

My main development machine is similar to your but weaker, and I can usually get between 30 and 200 pages per second with the tools I've been using (hand written and similar to Traffic Jamme.)

I see several questions raised by your post, but I am not sure which you are getting at.

AOLserver vs. Apache performance? If so, measure without the nsunix module in the way. Run Apachebench against the virtual server itself (have the virtual server implement nssock and give it a non port 80 port.)

nsunix performance? I've long been saying that I haven't seen the performance gains I would have thought should have been there when comparing nsvhr/nssock to nshvr/nsunix. But as I've said, I've usually done much better than what you're reporting.

Note too: you're testing adp pages with db queries. Get rid of the db queries and test again. Until you are benchmarking only apples, I am not sure what to make of your measurements....

Make sure all your setups are serving the same sized pages that use the similar (if not the same) underlying technology.

I am not sure about OpenACS, I have found that various ACS filters can quite dramatically slow down a site. Is that what you're getting at?

Uh, would you clarify your question?

3: Response to Server Performance Benchmarks with Apache Benchmark and Traffic Jamme (response to 1)

Posted by James Thornton on 04/30/01 05:27 AM

The Traffic Jamme result was from testing adp pages containing SQL queries, but the Apache Benchmark results were from testing AOLserver with the default "home" page that comes with Apache.

I used this page for my AOLserver sites because I came across a thread in the forum showing other people's results with it and I needed a standard (it contains a GIF and a PNG image). I'll run some more tests later, without nsunix.