Presenting Storage to Hosts

Presenting storage to Hyper-V hosts, which in turn make storage available to VMs, is no different than presenting storage to Windows Server that does not use Hyper-V. The considerations for Hyper-V revolve around how presented storage is configured in Windows Server.

Hyper-V has two main methods to present storage for use with VMs: Cluster Shared Volumes (CSVs) and Storage Spaces Direct (S2D). Both require a WSFC. S2D is an OS-based software-defined storage platform. It uses OS-level storage replication to make copies of data on locally-attached storage to other hosts in a host cluster. Storage only presented to a single host (“locally attached storage”) should not be used for Hyper-V implementations when FlashArray is involved.

For FlashArray, implementations of Hyper-V should consider CSV as well as a Scale Out File Server (SOFS) to store the virtual disks used by VMs. Pure published a document on SOFS configuration Scale-Out File Services (SoFS).

Both features rely on a WSFC. A CSV enables multiple nodes of a Hyper-V cluster to have concurrent read and write access to the underlying Volume(s). With CSVs, the storage can be owned by any host participating in the WSFC and can be failed over to another host without disrupting a VM’s connectivity to its storage or overall availability. Figure 9 shows a configured CSV.

Figure 9. Cluster Shared Volume in Failover Cluster Manager

How many CSVs are required? The answer is “it depends”. Configure as many as necessary based on your workload criteria and requirements. Create volumes on FlashArray and place your workloads carefully. Common considerations include determining which VMs that are required to be able to have snapshots and/or be replicated at the same time, running too many distinct workloads on a given volume which could cause conflicts and I/O Volume-level congestion within the hypervisor, different replication targets, and more. Always baseline and benchmark workloads to understand the performance footprint a VM will need.

Consider enabling the CSV Cache to improve read performance for commonly accessed data. Further details on the CSV Cache are available at the article “How to Enable CSV Cache” in Microsoft Learn. Microsoft’s documentation for CSVs can be found in the Learn article “Use Cluster Shared Volumes in a failover cluster” and with their free training “Introduction to Cluster Shared Volumes”.

Pure’s documentation on how to properly configure a CSV with FlashArray can be found in the article “Working with Clustered Shared Volumes on a Windows Server Host”.