vNext Technical Preview – Crash Prevention

The next version of Microsoft Windows Server and Hyper-V has planned support for additional resiliencies when storage failures are introduced to the environment. Should a virtual machine (VM) detect the underlying storage is no longer available, it will place it’s self in a paused state until access to the storage is restored as opposed to the operating system crashing. Having a storage failure is one of the worst types of failures an IT administrator can be faced with. Depending on the size of your organization, it can have a significant impact on your overall application up-time.

In our lab environment, I’ve bared witness to some very brief interruptions to storage availability and watched VM’s survive the interruption. Essentially, the VM stops writing to disk and will momentarily hang. Once storage becomes available again, the VM will resume writing to disk without the VM crashing. An extended disconnect (minutes) will generally result in the VM going down unless there is almost no disk activity.

In our production environment, Enseva leverages redundant Scale-Out File Server heads that connect to custom active-active storage arrays that do all kinds of fancy-pants performance acceleration (I’ll talk about that in other blogs). Due to the fact that customer data is stored on multiple active storage arrays and the Hyper-V nodes have four network paths to storage, it makes our cloud storage incredibly resilient. It’s still nice to see that Microsoft is adding features that could further reduce outage exposure in cases of user error, etc…