Hi Roy,

I agree on the RAID stuff, but you know how it works. DBA's are barely involved from the project's kickoff. The server was already set up with this RAID 5 disks group even before I had a chance to give my opinion. There are still battles to win before we eventually win the war (cf http://miracleas.com/BAARF/) 😉

One hour ago the distributor failed exceedeing the -QueryTimeout. It hasn't been able to read most of the MSrepl_commands table in 10 hours.

And the IO subsystem lead is quite confusing: the RAID 5 is able to deliver some 80MB/s whenever a checkpoint occurs, and not more than 4,3 Gb/s when the distrib.exe tries to read MSrepl_commands.

This server looks like an idle system.

I still have no news from the system engineer about this cache failure (FYI the event was visible until last week, not only right now, by the time the latency started to grow).

I'll try to post an answer when I have news from the SE. Thanks for your interest.

David B.