As we expanded the users on our GW7 & GW8 systems we were starting to get more and more "pending Client server" requests and eventually database rebuilds as the disk I/O was not able to keep with the demands of the NetWare 6.5 based POA.

We managed to determine the problem and what we discovered does not appear to be documented anywhere else.

Whilst monitoring the SAN I/O we were seeing around 80MB/sec of reads to the GroupWise server which seemed very high.

We decided to change the read-ahead setting on the NSS volume from the default of 2 to 64 to see if that would help. What we found was that I/O requests to the SAN increased dramatically and C/S started to increase and response slowed down.

Thus we then tried setting the NSS read-ahead setting to 0 (disabling read ahead) and I/O dropped to almost zero and performance increased dramatically!

We surmise that (at least in our SAN environment) disabling the read-ahead (setting it to zero) on NSS volumes dramatically improves the performance of GroupWise by eliminating a massive amount of apparently redundant reads. i.e. It may be worth checking as we all know GroupWise is a I/O hog.