MIRRORING THROTTLES BACKLOG

For some reason, the title of this blog post makes me grin because it evokes a mental image of one guy trying to comically strangle another guy.  But I digress…

Speaking at a user group meeting recently, an attendee asked whether the mirroring process had any capability to speed up or slow down its workload according to the amount of work waiting on it.  I cast around for an answer to this and my friend, Kalen Delaney, already had the answer.  She already has an answer for most any question on SQL Server, but again I digress.  She got her answer from Peter Byrne, a Microsoft program manager and member of the SQL Server Storage Engine team.

Peter says…”There is a throttling heuristic used by mirroring to try to keep the backlog from getting too large on the mirror.  Essentially, at commit time SQL Server may pause briefly if it determines the REDO queue on the mirror is getting large enough that failover time would be significantly affected.”

You can find more details about this process in the white paper:

http://www.microsoft.com/technet/prodtechnol/sql/2005/technologies/dbm_best_pract.mspx