We have started to migrate one of our reporting systems from Sql 2000 to Sql 2008 R2. One of the steps has been to test the perforance of certain Oracle linked server queries between each server. We are finding on average 3 fold better perforance
stats (in terms of query completion time) on the old server. This should obviously not be the case. The new server has signifantly more CPU/Memory/IO resources to play with, and it is 64 bit (not to mention its new!). Here's what I got
SQL 2000 on W2000 both fully patched. Old Dell Dual core with 3 GB of Ram running on two soft IDE Mirrored drives (yes I know... it sucks). It connecting via the MS OLE DB provider for Oracle (9i client)
ESX VMware Server with 2 CPU's assigned, 8 GB of ram connected to large HP SAN. CPU, ram and IO's have all been ruled out as the problem. We've tried varying network cards with different results so we havent ruled that out yet. Its connecting
via the Oracle provider for OLE DB (11G 64 bit client)
The linked server is an Oracle 9i fully patched server. All three are on the same network backbone.
Running a simple select * query on both servers returns the same number of rows (~76 000) . It takes ~1:20 on the new server and ~0:20 on the old server.
In looking at the wait stati
View Complete Post