June 4, 2009 at 3:09 pm
I've played around with performance monitor in the past, but I'm just now getting to the point where I'm ready to start utilizing it for diagnosing SQL Server issues, so I've got a lot to learn.
First of all, I've read that using the 'Batch Requests/sec' and 'Transactions/sec' counters will give a good indication of how busy my SQL Servers are. Does anyone have any other methods of doing this? I'm mainly trying to find the amount of Inserts, Updates, and Deletes per minute.
I guess I'd have to either 1) leave perfmon running all day OR 2) run it for a certain amount of time starting at different times throughout the week, and then average those values. Any other suggestions?
Also, all our servers are virtualized, but I'm guessing that wouldn't affect the SQL counters.
On that note, does anyone know of any good articles/books about running SQL Server on a virtual box?...or good articles/books about using perfmon for DBA purposes
Thanks all!
June 10, 2010 at 10:10 am
Running SQL Server on a virtual box (particularly production) is not a good idea. You can get into real trouble. In the real world, none of our clients do it. They might run it on dev or test, but then there is no way to benchmark or predict behavior. Scalability is an issue, also replication places additional burden.
June 10, 2010 at 10:28 am
Counters can get slightly screwy in VMs if the environment, or host, is busy. I have tended to look at a few types of counters for baselining. Batches/sec, CPU%, disk Q, memory reads/sec, cache hit ratio. There are some good baselining articles, and I'd search that for ideas.
In terms of running the servers as VMs, be careful of IO requirements. That's typically where people get into issues first. If things are heavily loaded, they might be better as physical machines.
Viewing 3 posts - 1 through 2 (of 2 total)
You must be logged in to reply to this topic. Login to reply