Backup taking too long after migration to SQL 2008 R2

  • apat (4/24/2013)


    Cluster is made of Dell R815 server. With Intel x570 10GB NICs connected to Cisco Nexus 5596 Switch through a coper twinex cables. That is connected to EMC VNX 5500 SAN with 10GB fiber. Protocol throughout is iSCSI. SAN is made up of SSDs and capable of over 35000 IOPS/second.

    Have you verified that you are getting 10GB through to the SAN? It certainly doesn't sound like you are.

    Have you gotten your SAN admins involved to see what they can offer?

  • SQL Server 2008 R2 is not slower running backups. In all likelihood you have some configuration or hardware issue with your connection to the SAN. Have you looked at wait stats from inside SQL Server?

    If you run queries to this disk, or copy data there, can you get an expected rate of transfer?

    http://blog.sqlauthority.com/2011/02/14/sql-server-backupio-backupbuffer-wait-type-day-14-of-28/

  • Infrastructure team is looking at the hardware configuration to find any possible issues there. While backup is running I see BACKUPIO and BACKUPBUFFER wait types.

  • apat (4/24/2013)


    Infrastructure team is looking at the hardware configuration to find any possible issues there. While backup is running I see BACKUPIO and BACKUPBUFFER wait types.

    Those both point to the IO system not being able to keep up.

  • Is power scheme high performance?

    I have had problems with balanced

  • Infrastructure team discovered a SAN issue. It is performing very slow and hardly 10 MB/s reads. Thanks for all your help, I will update the post with the resolution soon.

  • Let us know

    Thank u

  • change was made to removing native Windows MPIO and installing EMC's Powerpath management software, MPIO was not handling I/O properly and some of the paths to the storage were not utilized resulting in increased latency.

    Increased the Receive/Transmission Scaling Size window on the Intel 10G cards from 725 to 2048, increasing this setting helps when attempting to send large I/O blocks over the network.

    Backup size of 500 GB database came down from 6 hours to 2.40 hours. 1:30 is the actual backup and remaining time for verifying the backup file.

  • Still not the 40 minutes you claimed to get on the old server...... but 1:30 for 500GB is pretty decent.

  • yes still working to get close to 40 mins. I think that's the trade off between local backup Vs on SAN.

  • i think you would find that a properly configured SAN will almost always be faster than local disks. of course, your mileage will vary.

    you could have 8 internal SSD drives with a high capicity RAID card running RAID 10 that will probably be faster than a SAN.

    all SAN admins are not created equal... 😀

Viewing 11 posts - 16 through 25 (of 25 total)

You must be logged in to reply to this topic. Login to reply