I heard about maximum 50 Mb/sec for receive and 50Mb/sec for redo. But really, I have only 20Mb/sec redo speed. This is speed from Perfmon, so I know about the following issue:

As I know redo operation depends on CPU and for SQL Server 2012 it's run use only one core of CPU. I think it’s awful for situation when I can generate a lot of logs. For ex. when reorganize a lot of indexes in big tables.

Do you know about any KB or advise for improving performance for this case?


"Heard about"? From where? As far as I know this is dependent on local factors such as raw disk subsystem capability, controller capacity, concurrency, type of data being logged, etc. etc.

Redoing large index maintenance operations on AGs is definitely a known issue. This post certainly admits as much and suggests ways to mitigate the issues:

But I don't think you can expect any predictable x MB/sec never mind a guarantee, you will have to observe what your system is capable of, and troubleshoot issues when it falls below your threshold range.

  • In addition to what Aaron has said, unless you have 10 Gbit dedicated same data center + memory based (ssd/fusion io/violin) storage - I highly doubt you'll hit the max which is around 200 MB/sec in 2012. You'll see upcoming log transport fixes and much better throughput in 2016. Also, unless specifically turned off the log is compressed before sending so saying 20MB/sec of log sent is not completely accurate. – Sean Gallardy Feb 15 '16 at 20:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.