Re: Moving to a faster machine does not improve performance

[James Day]

Take a look at SHOW INNODB STATUS while hte system is under normal load. What are connections waiting for? Are the queries using the indexes you expect?

If you find that there are many daemons after the same records or index blocks you might consider trying several queues, perhaps for different types of work, and having the slaves poll each in turn.

If there is a single queue, you might consider using a queue daemon which responds to requests from slaves and updates the database itself. Atlernatively, there are some non-database functions in MySQL which may be of use - IS_FREE_LOCK, GET_LOCK and friends. If your applications are nicely cooperating you can use those to dodge the database storage engine without having to switch to a daemon. If you have simple task identifiers it may be relatively easy to use those to try to grab a lock on a task before switching to the storage engine for the final part of the status change. May or may not be less work for you.