Monitoring Databases in a Virtual Environment
The main points that I pulled from this session were
1. Task manager is a dirty filthy liar. Not a whole lot of specifics on this but it was a beginner session.
2. Don’t waste my time in SQL and check vmware ready time first, then go back to SQL.
3. DBAs should have read only access to vSphere to check host and guest counters.
4. Let vmware admins give you a 1vCPU box. Chances are they are faster CPUs and its easy to scale up later.
This was a great beginner session for DBAs new to vmware or looking to virtualize. My one comment is that he seemed to be stuck in a high gear when trying to get up to speed. Or another metaphor would be stuck at 60,000 ft when I would have liked to dive in a little sooner/faster.
Waits and Queues and You
Thomas described the running-waiting-runnable queue quite well including his salty humor. This queue is a first in first out system by design. The order of talking about queues first and then waits was good because the runable status in the queue is the one where the waits are tracked.
He again asked a question and told everyone who raised their hand that they were wasting their time. Note to self: don’t answer Thomas’s questions, however he would probably make a mockery of anyone not raising their hand. The question asked was if anyone was using sp_who regularly. I suppose this was a good lead in for waits that are in the DMOs.
He covered the big 4 CPU/NETWORK/DISK/MEM bottlenecks and what waits to look for when trying to diagnose which one of these is the culprit. I liked the fact that he went over his top 5 waits specifically. He had a demo script that showed how easy it is to query the DMOs and get wait information.