Design Considerations for Storage Spaces Direct

Sign in to queue

Description

Join Elden Christensen, Principal PM Manager in the High Availability & Storage team, to learn about the key design considerations for deploying Storage Spaces Direct in Windows Server 2016, including choosing drives and volume resiliency types.
 
To learn more, check out our docs on:

Or visit aka.ms/s2d for more!

Embed

Download

The Discussion

  • User profile image
    TolliL

    Good video - well explained. Especially the part about how many SSD may be faster flat than many ssd fronted by a few NVMe (depends on # and size).

  • User profile image
    Matthew Monday

    Hello Cosmos,

    I have a quick question about S2D. Is it possible to set up S2D with two storage pools? The reason I would need this is I am setting up cluster to cluster Storage Replica with S2D running within each of my 3 node clusters. I need to have a log disk run strictly on SSD and not be used for any other workload per Storage Replica requirements. This doesn't seem possible to me using S2D as I have no way of pinning a virtual disk to the SSDs.

    Thanks!

    Matt

  • User profile image
    cosmosdarwin

    @Matthew Monday: We strongly recommend using just one storage pool. (E.g. In the presence of multiple pools, features like auto-pooling automatically disable themselves.) I am guessing you have SSD and HDD, and the SSD will be used for Storage Spaces Direct caching. I just asked Ned Pyle, the SR PM, and he tells me that you can create the log virtual disk on that same pool (HDD with SSD cache) and it's absolutely supported. The only caveat is, obviously all-flash would be faster. :) I hope this helps!

  • User profile image
    Matthew Monday

    Thanks for the quick reply Cosmos!

    We initially set it up that way slicing the log volume out of our hybrid S2D array (SSD + HDD as you guessed), but I kept second guessing that decision and kept searching for a way to pin it to flash.

    I understand why you recommend one pool. It would be easy for someone to make a complete mess of this setup from powershell thus leading to alot of support calls for unsupported configs :). Enable-ClusterStorageSpacesDirect automatically configures this pool for you without any additional input. Aside from actually enabling S2D, this command must tag disks as cache or capacity and set up fault domains/enclosure awareness for the disks. Out of curiosity, what else would I need to do if I decided to autoconfig=false? Are there technical reasons a second pool is not recommended other than the unfriendly setup process?

    We will definitely be sending out the supported configuration you mention above. 2 clusters, 1 S2D pool on each cluster, and the log and data volumes on each cluster carved directly out of the S2D pool.

    In my test lab on the other hand, Ill be exploring S2D and SR further as I find both very valuable and interesting technologies. I find doing a little manual setup allows me to further understand how both work.

    Thanks!

    Matthew

  • User profile image
    James

    I too would love a reply to this question - I have a need to guarantee some VM's live on SSD storage, regardless of the systems assessment of if it counts as hot data or not..

Add Your 2 Cents