Cat 500 SQL Performance Issues

ben.dailey · ‎09-02-2007

This is a weird one. I have three switches:

1. Catalyst 2950 (100 meg with 1000Base-SX uplinks)

2. Catalyst 500 (100 meg with 1000Base-SX SFPs)

3. Catalyst 500 (gigabit)

Switches 1 & 2 are uplinked via 1000Base-SX, switches 2 & 3 are uplinked via 1000Base-T

The server farm (including a dedicated MS SQL server) is connected to switch 3.

Workstations connected to switch 1 experience normal connectivity to the SQL server but when connected to switch 2 are experiencing slow performance. It only appears to be with SQL server packets as other application level packets (e.g. SMB) appear OK. Lower layer connectivity appears to be working fine. Smartports on both switches are configured as workstations.

Can anyone shed any light on why SQL server connectivity from switch 2 is so slow?

Thanks.

ben.dailey · ‎09-05-2007

Still having problems with this - can anyone help please?

Joseph W. Doherty · ‎09-06-2007

"This is a weird one."

Agree, since network equipment by default is usually insensitive to applications. However, since the problem arises with the network location, unlikely its an App issue.

If you're only comparing SMB to SQL, I would think the latter might be a bit more chatty and use, on average, smaller packets. So, if there was additional latency or higher packet loss, the effect might be more noticeable to a SQL app.

If you haven't already done so, review all stats along the poorer performing path. Then you could either do some packet sniffing and/or additional synthetic performance testing.