Regular |
|
Joined: Tue Sep 30, 2003 11:27 am Posts: 60 Location: Columbus, OH, USA
|
Why can't you simulate 250 users with 100 million test records to find out? This should be a straightforward task for something like JMeter or Mercury Interactive tools, and you'd be well served to run a full PoC load test before taking anyone else's word for it.
There simply are too many variables to guess at otherwise (these questions are rhetorical):
* What are the details of your server and network architecture?
* What do your top 10 queries look like and how long do they take to run?
* What kinds of business logic must execute between the requests and responses?
* Do you have any syncronized or locking operations?
* Are you interfacing with any other systems? If so, in what part of the request process are they accessed?
* What's your user load profile look like?
And this is just scraping the surface. Seriously, set up some load tests and judge for yourself end-to-end, and bring in JBoss if you really want to cover your bases.
Scott
|
|