Virsto Creates VMware Storage Hypervisor to Boost VM Density and Performance, Interview with Virsto CEO Davis Part I

We have seen a lot of focus in the past twelve to 24 months on enterprise virtualization, virtual desktop infrastructure (VDI), and technology investments in storage. All of this emphasis on moving away from the traditional server-based deployments in the data center is putting a major emphasis on virtual server performance and, more specifically, on enterprise storage performance. Today, I begin an interview series with Virsto Software CEO Mark Davis, where we look at how Virsto creates a VMware storage hypervisor in VMware vSphere to give incredible boosts in performance using even traditional hardware.

Ben: Virsto is doing some incredible things by simply integrating into the hypervisor. In reading over the Virsto documentation, I see that your product is trying to move toward a storage-focused hypervisor.  Could you talk to me a little bit about what exactly Virsto is doing?

Mark: Virsto has been using the word “storage hypervisor” for what we do for some time. We think that a storage hypervisor needs to have three fundamental characteristics.

  • One is that it is a completely virtual machine-centric approach to storage. We think that traditional storage-centric methods and architectures are inappropriate for virtualization and cloud computing. We think that a 100% software VM-centric way of building the storage hypervisor creates a lot of value for customers in the same way that a completely software-oriented server hypervisor did the server world.

  • The second thing is that a true storage hypervisor needs to provide a new kind of virtual storage object that is managed and consumed in the life cycle of virtualization. This new type of storage object should have some very different characteristics from what we are used to. It should always be high performance.

We should be able to deliver a consistent quality of service out of that storage object no matter how it is used or configured. We should not get high performance sometimes and low performance others, which is what we suffer from today in virtualization.

The virtual storage object should always be space efficient in addition to always being high performance. By space efficient we mean a couple of key things.

One is that it is always thin provisioned. We would not think of virtualizing without thin provisioning – virtualizing our servers is thin provisioning our servers. We need to be able to oversubscribe storage.

We have to be able to do it all the time, not just when the performance requirements are lower and not just when I can afford to buy a super high-end disk array. Every virtual machine storage object ought to be thin provisioned with no compromise in performance.

They also should be able to make these storage objects, Virtual Machine Disks (VMDK) and Virtual Hard Disks (VHD), out of completely scalable snapshots and clones that always share the common blocks that are shared amongst all of the virtual machines that are running, so that we do not unnecessarily copy data ever.

  • The third thing a storage hypervisor needs to do is deliver a radically new economic model for storage in the data center. It should support any block level storage, it should support any vendor, and it should support storage that might be very high end. It should also support storage that is much more commodity-like. It has to be multi-hypervisor so that organizations are not locked into a particular hypervisor vendor. It should also be designed and architected for not medium size but truly cloud-scale deployments.

So those are the three characteristics of the hypervisor we think of:  The virtual storage approach, the new type of virtual storage object, and the new storage economics.

Ben:  Is Virsto virtualizing what we would normally think of as a block device? Or is it something else?

Mark: Virsto virtualizes what the virtual machine administrators manage, which is specifically in the case of VMware, called the VMDK. In the case of Xen or Microsoft Hyper-V, it is called a VHD or virtual hard disk. These are kind of encapsulated storage objects that are actually to a guest operating system looks like a disk volume.

That is what we mean by this kind of object. Virsto calls it a “vDisk” because we do this across multiple platforms with different names. Our vDisk, to the hypervisor or to the guest operating system, looks, smells and taste like a VHD or a VMDK, depending on the platform.

They are indistinguishable from those native objects, from the point of view of how they are addressed, the name space, and the like. The big difference is they are always space efficient, they are always thin provisioned, and they are always very high performance, unlike the native implementation.

In Part II of my series with Virsto CEO Mark Davis, we will take a look at how Virsto works with virtual machine hypervisors to make device I/O  up to ten ( 10) times faster than the traditional throughput seen in today’s data center hardware.

In Part III of this interview series, we talk about where Virsto sits in the vSphere stack and how it works to deliver these increases in performance.

In Part IV of my blog series with Virsto Software CEO Mark Davis, we will look at how Virsto fits into the private cloud infrastructure storage space and what it does to optimize the performance of SSDs.

Click Here to Signup for the DCIG Newsletter!


DCIG Newsletter Signup

Thank you for your interest in DCIG research and analysis.

Please sign up for the free DCIG Newsletter to have new analysis delivered to your inbox each week.