I like this article. Even though you say it only touches the surface of the subject it goes into it enough to give me a better understanding of stored procedure caching.

I have a couple questions about portions of your article.

First, when you say that a stored procedure becomes serialized when being compiled and as such can cause bottlenecks, do you mean that while being compiled no one can execute it until the compile is finished or does this just mean that it can't be modified until the compile is complete?

Second, I tried a test of running a stored procedure not in the master that started with sp_ and the same one in the same database that didn't start with sp_. The difference in milliseconds was very small. These were simple stored procedure. Would the benefit in speed be greater with a more complex stored procedure?

Robert Marda