Disk IO and Version 9

Question

We recently upgraded from 8.5 to 9.1, and saw our disk IO shoot through the roof for our database drive. We were seeing the Average Disk Que Length on the drive stay under 1 with version 8.5, except during database maintenance. We are now averaging 10 all the time. We made no other changes to our system.

We run the database on its own server, SQL 2005, Dual Quad Xeons, 4GB RAM, 5+1 Dedicated HDDs for the database in a RAID 5.

I have opened a ticket with support. Anyone else seeing this?

FormerMember · Answer

In order to better support the product, we have added additional logging, which results in more database activity.  We're going to turn the volume down in SP3.

NasaGeek · Answer

Turning on Adaptive Read Ahead has brought the IO back to nominal levels. However, this begs the question, why is Orion performing so much more reading with 9.1 then 8.5?

NasaGeek · Answer

After looking at the issue again, I noticed that we were not losing any data points. I looked further into the disk IO and found that the high average count is related to reads, not writes as I first assumed. The write ADQL on this disk is .09, and this is why we are not losing data points.

I do not have 'read ahead' turned on controller because previously it was not needed, but I will turn it on later this week and see if the IO returns to a nominal state. I will post the results.