So I started to do some research on the issue that I was looking at yesterday (By the way I believe we are still seeing the problem but has not been validated by the client). So this all got me thinking, I needed to take a new approach, Interesting thing here is I pulled from 3 sources none of them Books on Line. See the though process that started all this was a server upgrade. As I mentioned earlier there is a lot of Memory problems on this server. But the big thing is that I failed to mention before the migration we were on a 32 bit system and we upgraded it to a 64 bit system. Now here is where I am not sure if I understand all this and I am not sure how to approach this. Let’s start by the server configuration.
- 64 bit server
- 8 Gigs of memory
- AWE switch is on
- SQL Server Enterprise
- SQL Server is Clustered
- Windows 2003 Enterprise.
Ok so here is the big question…
If we are on a 64 bit system and if I remember off the top of my head it will support 32 gig’s of memory. Rather then the 2 that is with the 32 bit system. Then why did the hosting company configure it to use AWE memory? Is it needed, SQL Server will use it if it see it. Now I know there is more behind the AWE option then just using it. But here is what I found.
- 1 article from SQL Server MVP says that you should not turn it on, it increases the amount of overhead that you do not need.
- 1 blog from a Microsoft SQL Server engineer says leave it on.
- 1 Article says leave it on but then give only 75% of the memory to SQL Server because the OS will need the rest.
I find that there is not as much information on how to monitor SQL Servers memory, and explain how SQL Server uses its memory. I am going to try to get in touch with a couple people that I know today, but I think my recommendation at this point is to turn off AWE for 24 hours and see how it runs. This goes back to the baseline that I was talking about. This way I can see if it made a positive difference, negative difference, or if it was no different.
At this point I can understand the clients concern. I shared with them that if there is a code problem upgrading to a bigger and better server may not help at all. At best it will speed the system up but when the load increases in proportion to the system then we may see the problem again.
I am interested in what you think. If you have time drop a comment in here for me.