cya all untill 9.3 i guess.

The Common Test “server” is not something you’d really use for performance and load testing, except perhaps to a very limited extent. What you want to do is develop automated systems to put extreme loads on particular subsystems (servers and server clusters) that simulate high production loads, and see how they fail. For example, if you’re running a large “web server” (which will often be a whole set of servers serving different parts of the the content, such as one set for images, a set of caching servers for content that doesn’t change much, and a set of application servers for user-specific pages) you might use something like ApacheBench or, much better yet, httperf, to generate a load that will find the bottlenecks that can bring a system to its knees, and then have your programmers work on the problems that cause the system to collapse.

That’s the quick, and not terribly precise, summary. Anybody who’s got thoughts on this, and hasn’t run, say, httperf against a web app he’s written, should do so first to gain the understanding of what really goes in to this sort of thing.

WG certainly has the resources to hire people who are good at this sort of thing. That does not, however, mean that their company culture would actually allow them to do so.

Remember, the test is only up a few times, every 1-2 months. That’s a LOT of rather expensive hardware to have sitting around taking space.

bad bad bad…