FsLogix Storage setup with Nutanix RRS feed

  • Question

  • Hi,

    We are using FsLogix (Profile container and Office container enabled) on Windows Server 2008 R2 RDSH server.  Currently we have around 10 RDSH servers (no Citrix) in the Farm with maximum of 650 concurrent logons.  We have enabled concurrent User Profile Access on FsLogix Profile container.

    The FsLogix storage is on Nutanix with 2.5 TB single disk, which is presented to Windows Server 2016 File cluster with iSCSI connection.  This storage is used for both Profile and Office container.

    During peak times we have been experiencing Server Freezing/Hangs.  We were able to observe the file server reporting high disk queue length during the peak time.

    We have engaged Nutanix support to investigate the storage performance.  Below is their findings and recommendation:

    1 - Write I/O on the volume used by FsLogix is almost 100% in nature.

         This means that almost all of it is ingested by our oplog once it lands on the storage.

         Moreover, the working set on this volume is very high, with around 25 GB written every two minutes

         All of this means that the oplog itself, which has a 6 GB limit per vdisk, gets saturated very quickly.

         Once full, emergency flush operations start, but these temporarily halt the I/O on the vdisk, causing performance issues.

    2 - A lot of I/O coming to the vdisk is misaligned, which can have a very high performance impact (


    1 - For the first point, my recommendation is to split the vdisk.

         At the very least try to use 5*500 GB vdisk, but, given the working set, going up to 10*250GB vdisks may prove beneficial.

         Please check if FsLogix allows the use of more volumes to span the vhdx profiles, otherwise Storage Spaces ( could be leveraged to create a bigger volume

     We are planning on creating multiple disk on Nutanix and present to Server 2016 to create storage pool cluster volume.

    My question is, will FsLogix have any issue using Storage pool Cluster volume?

    Also we are also planning to format the disk in REFS instead of NTFS.  Is it true using REFS will help speeding up the logoff since we use concurrent user profile access?

    Thank you for any advice/recommendation regarding this.


    Friday, August 9, 2019 11:13 PM

All replies

  • Hello,

    If I am reading this correctly you are running on one actual disk. That certainly is probably overload for one disk. I would split the load among multiple disks.

    In regards to REFS, unless the whole chain (Client OS and storage) is at the level which supports complete REFS functionality, I don't believe you will gain substantial benefits from REFS. But you can try it. There are a lot of different variables here so we can't say for sure what exactly you will see in your environment.

    As far as the back end storage (storage pool cluster volume or other), FSLogix doesn't have any specific requirements.


    Monday, August 12, 2019 5:17 PM
  • BTW, you can also see the forum thread Storage Recommendations for FSLogix for general information on storage with FSLogix.
    Monday, August 12, 2019 5:22 PM
  • Thank you Micah,

    We have just started testing the new environment, will update our test out come here.



    Tuesday, August 13, 2019 7:52 AM
  • Did you get to any conclusions regarding the issues that you saw? We've seen similar and we're using FSLogix Profile Containers + Office365 Containters.

    How are you compressing the storage containers which holds the FSLogix disks etc?
    Monday, February 17, 2020 5:42 PM