My SQL Server is a virtual machine, running on ESX 4.1 and configured as follows:
- 28GB of Memory, 21500 configured in Max Server Memory
- Reservation is set in vcenter to 28GB, so there should be no balloon driver issues
- SQL Service is configured for lock pages in memory
- According to sys.dm_os_performance_counters, Target Server Memory and Total Server Memory are equal, at 537600KB, i.e. approximately 5GB
- Optimize for Ad Hoc Workflows is enabled
I know that the Page Life Expectancy value itself isn't necessarily important, but rather the overall PLE trend. We've had sudden drops in it lately during periods of heavy use of one of our web sites using the server as a backend. During the drops, performance in one of our critical web applications slows to a crawl. What I don't understand, is why SQL is only using 5GB of memory if it has 21.5GB available to it. Shouldn't it be caching as much as possible in the buffer pool? It was my understanding that SQL would take as much memory as it could, but this doesn't seem to be the case. The general consensus on the net seems to be that adding memory to a server is a quick and dirty way of working around PLE issues caused by poorly written queries, but if the server is only using a fraction of the memory available to it, why would increasing the memory help?