Hi Tyler,
this is indeed a very important effort. I do occasionally go through
different kind of pages on the web applications that we develop and
profile them with Yourkit while hitting them as hard as possible with
Siege.
These are the areas that I identified recently as 'interesting' to
profile:
* framework overhead (simple hello world)
* stateless embedded elements in a single page tree structure
* stateful embedded elements in a single page tree structure (both
with session state storage and client state storage)
* flow execution and data transfer of a multi page sub-site
* 'AOP' overhead of behavioral inheritance (ie. the inherits keyword
- sometimes referred to as 3D flow)
* continuations overhead in a single checkout element
* performance of clearly identifiable CRUD operation (this could just
benchmark RIFE/Crud in fact) - How many data items should be added to
the DB for this to be useful? Ten thousand?
CRUD does most of the basic operations, bean population, validation,
error reporting, constraints, GQM usage and ContentQueryManager
usage, ...
Anyone else have some insights on how to proceed with this? I have a
free license of Yourkit that can be used by anyone that can be used
for RIFE development (same for free licenses of IDEA). If anyone
wants these to work on RIFE, let me know.
Best regards,
Geert
I've been pondering the idea of working up some "official"
benchmarks for RIFE and I wanted to get everyone's input. Basically,
how should the test be designed. I think the data would be helpful to
determine any hotspots in RIFE and pinpoint RIFE's strengths /
weakness, and scalability.
1) What does everyone consider an average page on RIFE (database
access, template generation, CMF access, etc.)
Obviously, template generation is important, but I'm iffy on the
Database access. The only way RIFE really "changes" normal DB access
is it's SQL generation, but I guess that could be a hotspot and that
is definitely something most of us use, so maybe it should be
included.
2) How complex should the test be? (related to #1)
Should we include Bean Marshaling? (aka. Form Post-back?), page
redirection, database access, etc.
3) Should there be multiple "average" pages?
Do we make a simple template generation page, a database access page,
a bean marshaling page, an everything page, etc.?
4) What should we use to stress test the framework? Siege, something
else, something custom.
Personally, I think using an industry standard like Siege is probably
the best idea.
Cheers,
Tyler
_______________________________________________
Rife-users mailing list
[email protected]
http://lists.uwyn.com/mailman/listinfo/rife-users
--
Geert Bevin
Uwyn "Use what you need" - http://uwyn.com
RIFE Java application framework - http://rifers.org
Music and words - http://gbevin.com
_______________________________________________
Rife-users mailing list
[email protected]
http://lists.uwyn.com/mailman/listinfo/rife-users
