none
Server 2012 iSCSI Target in Production Hyper-V Cluster? RRS feed

  • Question

  • I am looking to update our Hyper-V cluster and am wondering about using the server 2012 iSCSI target as my shared storage between 3 hosts.  Does anyone know if the iSCSI target is ready performance and stability wise to be used for multiple VM workloads including SQL?

    I have a server I would re-purpose for this if so:

    3ghz Xeon quad core
    8GB RAM
    10x 300GB 15k SAS drive

    Also if you think this would work, would it be best to RAID the disks with the hardware controller and then share via the iSCSI Target or better to JBOD the disks and use Storage Spaces for the "RAID"?


    My hosts and this server will of course have dedicated GBE for iSCSI and a separate switch.
    • Edited by Thildemar Monday, August 26, 2013 9:33 PM
    Monday, August 26, 2013 9:32 PM

Answers

  • I am looking to update our Hyper-V cluster and am wondering about using the server 2012 iSCSI target as my shared storage between 3 hosts.  Does anyone know if the iSCSI target is ready performance and stability wise to be used for multiple VM workloads including SQL?

    I have a server I would re-purpose for this if so:

    3ghz Xeon quad core
    8GB RAM
    10x 300GB 15k SAS drive

    Also if you think this would work, would it be best to RAID the disks with the hardware controller and then share via the iSCSI Target or better to JBOD the disks and use Storage Spaces for the "RAID"?


    My hosts and this server will of course have dedicated GBE for iSCSI and a separate switch.

    Microsoft iSCSI target is a very stable and solid solution for test & development however with Windows Server 2012 SMB 3.0 will be faster. For a simple reason: Microsoft iSCSI is not cached on the server side and Microsoft SMB 3.0 redirector *IS*. So if you're fine with your storage being a single point of failure then deploying SMB 3.0 is recommended over *Microsoft* iSCSI (other commercial implementations are faster, say StarWind does caching and DataCore does tiering with flash and both can support multiple nodes so do active-active and Microsoft is active-passive only so does not scale with multiple instances of target added to the cluster). If you want to make SMB 3.0 fault tolerant you may either create an SMB failover cluster or deploy SoFS (depends on your workload, SoFS does not support generic worload unfortunately...). See:

    http://technet.microsoft.com/en-us/library/hh831349.aspx

    SAS JBOD require an external enclosure (better two if you want redundancy and fault tolerance) and whole bunch of the corresponding SAS infrastructure like SAS disks and controllers. See:

    http://blogs.msdn.com/b/clustering/archive/2012/06/02/10314262.aspx

    Alternative way would be to spawn a software doing synchronous mirroring between your hosts and creating fault tolerant high performance LUNs exposted to system. See sample here:

    http://www.starwindsoftware.com/native-san-for-hyper-v-free-edition

    This one can live with "el cheapo" SATA spindles and does not require anything except actual Hyper-V servers and network connection between them.

    Hope this help :)


    StarWind iSCSI SAN & NAS

    Tuesday, August 27, 2013 12:35 AM

All replies

  • I am looking to update our Hyper-V cluster and am wondering about using the server 2012 iSCSI target as my shared storage between 3 hosts.  Does anyone know if the iSCSI target is ready performance and stability wise to be used for multiple VM workloads including SQL?

    I have a server I would re-purpose for this if so:

    3ghz Xeon quad core
    8GB RAM
    10x 300GB 15k SAS drive

    Also if you think this would work, would it be best to RAID the disks with the hardware controller and then share via the iSCSI Target or better to JBOD the disks and use Storage Spaces for the "RAID"?


    My hosts and this server will of course have dedicated GBE for iSCSI and a separate switch.

    Microsoft iSCSI target is a very stable and solid solution for test & development however with Windows Server 2012 SMB 3.0 will be faster. For a simple reason: Microsoft iSCSI is not cached on the server side and Microsoft SMB 3.0 redirector *IS*. So if you're fine with your storage being a single point of failure then deploying SMB 3.0 is recommended over *Microsoft* iSCSI (other commercial implementations are faster, say StarWind does caching and DataCore does tiering with flash and both can support multiple nodes so do active-active and Microsoft is active-passive only so does not scale with multiple instances of target added to the cluster). If you want to make SMB 3.0 fault tolerant you may either create an SMB failover cluster or deploy SoFS (depends on your workload, SoFS does not support generic worload unfortunately...). See:

    http://technet.microsoft.com/en-us/library/hh831349.aspx

    SAS JBOD require an external enclosure (better two if you want redundancy and fault tolerance) and whole bunch of the corresponding SAS infrastructure like SAS disks and controllers. See:

    http://blogs.msdn.com/b/clustering/archive/2012/06/02/10314262.aspx

    Alternative way would be to spawn a software doing synchronous mirroring between your hosts and creating fault tolerant high performance LUNs exposted to system. See sample here:

    http://www.starwindsoftware.com/native-san-for-hyper-v-free-edition

    This one can live with "el cheapo" SATA spindles and does not require anything except actual Hyper-V servers and network connection between them.

    Hope this help :)


    StarWind iSCSI SAN & NAS

    Tuesday, August 27, 2013 12:35 AM
  • If there is anything further we could help please just let us know.

    TechNet Subscriber Support in forum |If you have any feedback on our support, please contact tnmff@microsoft.com.

    Wednesday, August 28, 2013 11:37 AM
    Moderator