I'm always surprised by those who find their first generation SOAs slow. When I dig into the architecture, and the process of getting their, I almost always find that there was little of no consideration of performance, no modeling nor testing.

Performance is a project killer, as we discovered over the years. If the SOA does not perform well, then those that leverage its edge applications, especially composites, won’t find the term "SOA" as something good. Indeed, many SOAs have gone back to the drawing board due to performance issues they did not expect, or were able to architect around.

So, how does one deal with the notion of performance with SOA? It's really a matter of check points, testing, and selecting the right enabling technology. However, much can be done using performance models, or mathematical models that will estimate the performance of your SOA before you build anything or purchase technology. You can find out all about performance modeling in this book, or just think in terms of queuing models you may have built in college. It's not new stuff.

SOAs are not unlike any other distributed computing systems, and thus designing a performance model should be nothing too new. At this point we understand exactly how each service behaves under an increasing load, and we have enough data to plug into a model. Now, it's just a matter of building a model. I'll pick up on that tomorrow.

Posted by Dave Linthicum on June 7, 2006 06:40 AM