Hi All,
We have a SQL Server 2008 Standard Edition virtualized with VMWare, running on an ESX host that has 2 sockets x 6 cores with Hyperthreading ON. That gives us 24 CPUs.
We have created that SQL Server VM and assigned it 24 CPUs.
Reading a MSFT best practices article on setting MAXDOP for servers using hyperthreading, the suggestion was to change MAXDOP to the actual number of cores. In our case it would be 12 (2 sockets x 6 cores).
My question is, if we change MAXDOP to 12 (from default 0), but SQL Server sees 24 logical CPUs, how does SQL Server know that when it parallelizes a query, that it spreads it among cores and not hyperthreaded CPUs?
By the way, i have checked WAITSTATS, and CX_PACKET is 70% of what SQL server is waiting on, so i figure changing MAXDOP might help.
Thanks in advance.
