Hi,

Just run some benchmarks against Tomcat 3.3-m1 :

Server was a Linux Redhat 6.2 box with PIII/800 + 256Mo.
128Mo were allocated to tomcat :

All tests conducted with ab :

ab -c 10 -n 1000 host

Test with Apache 1.3.17 and Apache 2.0.alpha12  
mod_jk log desactivated with JkLogLevel error instead of warn

Apache 1.3 serving index.html 
Requests per second:    192.64

Apache 2.0 serving index.html           
Requests per second:    180.08

Tomcat serving HelloWorld directly (no ajp)
Requests per second:    954.20

Tomcat serving HelloWorld via Apache 1.3 + mod_jk + ajp13 
Requests per second:    427.17  (cachesize = 1)
Requests per second:    481.46  (cachesize = 16)

Tomcat serving HelloWorld via Apache 2.0 + mod_jk + ajp13 
Requests per second:    504.54  (cachesize = 1)
Requests per second:    474.83  (cachesize = 16)

Tomcat serving HelloWorld via Apache 1.3 + mod_jk + ajp12
Requests per second:    9.63    (something bad there ?)

Tomcat serving HelloWorld via Apache 2.0 + mod_jk + ajp12
Requests per second:    24.00   (something bad there ?)

Example Servlet info :
Document Path:          /examples/servlet/HelloWorldExample
Document Length:        406 bytes
Concurrency Level:      10
Complete requests:      1000
Failed requests:        0
Total transferred:      645645 bytes
HTML transferred:       406406 bytes

Static info:

Document Path:          /index.html
Document Length:        1520 bytes
Concurrency Level:      10
Complete requests:      1000
Failed requests:        0
Total transferred:      2004000 bytes
HTML transferred:       1520000 bytes



Some conclusion :

- The Tomcat HTTP connector is the clear winner and if its optimisations
  are ported back to ajp we may see something good here

- Apache 2.0 is not faster than Apache 1.3 but it's still in Alpha release. 

- The c set to 16 didn't show any improvement but test must
  be reconducted with ab using -k flag (HTTP KeepAlive feature)

- ajp12 show a strange comportment and must be investigated further -(


Future benchs :

I plan to do the same tests but using LoadBalancing features of mod_jk 
against 3 tomcat running on same server configurations 
(Redhat 6.2, PIII/800, 256Mo, TC use 128mo)


Also use the HTTP KeepAlive feature to see if cachesize show improvment.

Regards

benchs.zip

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, email: [EMAIL PROTECTED]

Reply via email to