• Hi dma33 & John

    Linchi Shea has posted an analysis on the impact of page checksums and default trace with various IO workloads. You can review it here. I didn't want to go back there, as Linchi had already experimented and explained the most important facts. I just wanted to put the emphasis on the upgrade problem, and mention the readonly scenario.

    John, if you choose to deactivate page checksums, you won't be warned when there are torn pages and bits trashed by anything lying under the file system. It is much safer to let the parameter to ON, it is not just another BP. Ask Paul Randal or any PFE or escalation engineer who has worked on a corruption case, and check the Data Corruption section on this forum, these 823, 824 and 825 errors are happening.

    HTH

    David B.

    David B.