Some benchmark results....

I ran a quick mini-benchmark of a modified AddressBook WSIF sample, that 
compared invocation times for direct java invocation of the address book, 
WSIF without and with caching.

Running 10K iterations on a 3.2 ghz wintel box on JDK 1.4 I got the following 
results:

[java] Times in milliseconds...

[java] Direct Total:               94.0
[java] Direct Average:           0.0094

[java] WSIF NonCached Total:       117703.0
[java] WSIF NonCached Average:   11.7703
[java] WSIF NonCached Increase:  11.760900000000001

[java] WSIF Cached Total:          2687.0
[java] WSIF Cached Average:      0.2687
[java] WSIF Cached Increase:     0.2593


As you can see, non-cached WSIF invocation added about 12ms per invocation.  
Turning caching on added just over 1/4 millisecond.

Quite respectable!  And it means that injecting WSIF as a decoupling layer 
and thereby using WSDL as an interface definition language, thus allowing you 
to substitute real web services, BPEL engines, MOM, EJBs and the like at a 
later date with no or minimal code changes.

Just thought I would pass on the fruits of my labours! ;-)


Andrzej Jan Taramina
Chaeron Corporation: Enterprise System Solutions
http://www.chaeron.com