D
dbeavon
Guest
Right disk is one of my primary resources and will be a bottleneck from time to time. My goal would be to find a way to add that buffered data into memory on a more permanent basis (larger buffer pool, alternate pool, or whatever). Or to move the database off the current disk to a better one. Or "plan C" - use this test or something like it as a baseline to determine if our disk resource is possibly misconfigured. In the SQL Server world we have a tool called "sqlio" that is used ( independently of SQL Server itself ) to run performance tests against the underlying disk (various block sizes, queue lengths, read vs write preferences, random vs sequential, and so on). Not sure if Progress ships such a thing (eg. for HP-UX) to measure the performance characteristics of the underlying disk that it sits on. Before moving forward on any of these, we would need evidence that points conclusively to a 10 second disk bottleneck. A repeatable test that involves re-reading from disk would be the most straight-forward evidence.
Continue reading...
Continue reading...