Primary Storage

 View Only

  • 1.  Number of Vdisks considerations

    Posted Fri September 22, 2023 01:01 PM
    Edited by T Masteen Fri September 22, 2023 01:06 PM
    Hi All,
    In the context of SVC 8.5.x or 8.6.x in combination with ESXi 7.x, are there SVC performance reasons to create 2x 8TB vdisks instead of 1x 16TB vdisk as a data store?
    - We use regular pools with Flashsystems as backend
    - Several multi-host clusters
    - There will be many vdisks (independed of size) mapped to ESXi Host clusters
    - 4 and 6 Node Enhanced Stretched Clusters
    Thanks



    ------------------------------
    TMasteen
    ------------------------------



  • 2.  RE: Number of Vdisks considerations

    IBM Champion
    Posted Fri September 22, 2023 01:18 PM

    It seems unlikely that you're going to have much difference based on just the volumes. There are going to be things that you need to work out beyond that in terms of how the systems are laid out and mapped on the front and back of the SVC. But the difference between (1) 16 and (2) 8s from a performance perspective isn't going to be too big of a deal in general, I think.

     

    Regards,

    Ian R Wright

    Client Technical Sales Engineer

    Mainline Information Systems

    Specializing in Mainframe and Open Storage, Linux, Ansible, Containers/Cloud 

     

    ian.wright@mainline.com

    301-514-0758   

     

            

     

     

    This e-mail and files transmitted with it are confidential, and are intended solely for the use of the individual or entity to whom this e-mail is addressed. If you are not the intended recipient, or the employee or agent responsible to deliver it to the intended recipient, you are hereby notified that any dissemination, distribution or copying of this communication is strictly prohibited. If you are not one of the named recipient(s) or otherwise have reason to believe that you received this message in error, please immediately notify sender by e-mail, and destroy the original message.



    Original Message


  • 3.  RE: Number of Vdisks considerations

    IBM Champion
    Posted Mon September 25, 2023 02:50 AM

    Hi,

    performance is a point to look at from a vm perspective in a vmware environment rather than the esxi host itself. for example if your vm is a database server and uses multiple disks for it's database, log and index files, multiple datastores (multiple vdisks) performs better than the single vdisk - single datastore configuration. 



    ------------------------------
    Nezih Boyacioglu
    ------------------------------

    Original Message


  • 4.  RE: Number of Vdisks considerations

    Posted Wed September 27, 2023 11:04 AM

     Better resilience apart (a VMFS datastore can get corrupted as any other filesystem) there are also a performance reason: preferred node.

     The preferred node is the node that runs all write operations to the backend disks. Each volume has a preferred node and if you have only one volume, you can only use one node as preferred. With two volumes you can balance this function between to nodes.

     In a stretched topology, the way this impact in host's preformance depends of the ISLs speed and latency and the way the SVC infrastructure manage the writes (the mirror write priority) but it will be better balanced with two volumes than one.
     
     We have deployed several Enhanced Stretched Clusters and we try to balance the sites and nodes workloads as even as possible.

     Please, let us know what you have implemented!.



    ------------------------------
    Javier Gavilan Lopez
    ------------------------------

    Original Message


  • 5.  RE: Number of Vdisks considerations

    Posted Wed September 27, 2023 11:05 AM

     Better resilience apart (a VMFS datastore can get corrupted as any other filesystem) there are also a performance reason: preferred node.

     The preferred node is the node that runs all write operations to the backend disks. Each volume has a preferred node and if you have only one volume, you can only use one node as preferred. With two volumes you can balance this function between to nodes.

     In a stretched topology, the way this impact in host's preformance depends of the ISLs speed and latency and the way the SVC infrastructure manage the writes (the mirror write priority) but it will be better balanced with two volumes than one.
     
    We have deployed several Enhanced Stretched Clusters and we try to balance the sites and nodes workload as even as possible.

     Let us know what you have implemented!.



    ------------------------------
    Javier Gavilan Lopez
    ------------------------------

    Original Message


  • 6.  RE: Number of Vdisks considerations

    Posted Wed September 27, 2023 01:45 PM
    To give some more context to the question:
    In the Redbooks, it is indicated under Back-end configurations that the number of vdisks (and therefore the size) may be important.
    quote: For all-flash back-end arrays, a best practice is to create at least 32 logical volumes from the
    array capacity to keep the queue depths high enough and spread the work across the
    virtualizer resources.
     
    In the same Redbooks there is no mention of the number of Vdisks (or capacity) at SVC level towards hosts.


    ------------------------------
    TMasteen
    ------------------------------

    Original Message


  • 7.  RE: Number of Vdisks considerations

    Posted Fri September 29, 2023 08:20 AM
    Edited by Javier Gavilan Lopez Fri September 29, 2023 09:04 AM

    Hello. That's correct, but it doesn't refer to the volumes presented to the hosts. It does refer to the back-end storage volumes. In fact, if the back-end arrays are Flashsystems, they would auto create the volumes with the apporpiate size and node (back-end node) balancing if you select the "Automatic Configuration for Virtualization".

     But again, it does not refers to the hosts view of the storage.



    ------------------------------
    Javier Gavilan Lopez
    ------------------------------

    Original Message