VM-aware storage is a hot topic right now in the virtualization world, according to storage and virtualization expert Stephen Foskett. In the past year there was an increase in storage developed specifically for virtual machines. These storage systems enhance data migration, performance and integration with virtual server environments. However, some VM storage products will only work well if virtual machines are the only workloads running on it, which may not benefit organizations that haven't virtualized their entire environment. In this podcast, Foskett discusses the benefits and drawbacks of VM-aware storage, and where he sees the technology going in the future. Listen to the podcast or read the transcript below.
How are vendors making VM-aware storage different from traditional storage, and what are some of the common features that can be found with it?
Stephen Foskett: That's really a major industry trend right now. If you're going to start a storage company, virtualization-aware storage is one of the things you're going to focus on. That, or maybe people are doing mass and scale, and cloud, and service provider storage, and things like that, but really virtualization is one of those key ingredients. What they're doing -- and this is not just the startups, but the established vendors are working in this direction as well -- is increasing communication between the storage array and the hypervisor. [That way] the hypervisor can do what it does and allow the storage array to offload work like copying data and moving data, acceleration, integration with virtualization management features and things like that. It's really a lot about communication, and that's one of those things that's happening in storage right now. You look at VMware with its VAAI APIs that allow the hypervisor to tell the array "copy this data" or "clear this data that's thin provisioned" and things like that; that ends up being a real differentiator for these storage products. And arrays that have better data movement, and better performance, are much more compelling to somebody who is building a virtual infrastructure than a standard storage array.
You mentioned VM-aware storage having better data movement and performance. What are some of the other advantages to using it?
Foskett: The number one advantage is that … virtualization breaks storage. All of the assumptions that storage arrays make about server workloads are based on sort of a one-to-one mapping of application to server support, to connections, LUN, all that stuff. It's assumed by the array that [all the mapping] is a simple task. For example, an array can make a copy of a LUN on the assumption that the LUN is a specific server and application workload. But in a virtualization environment -- we call it the I/O blender -- what happens is that LUN may be part of the storage of a dozen or 100 servers and applications, and what that means is the array just can't do anything well. It can't accelerate performance, it can't copy data, it can't share data effectively, without knowing what is going on [with] the storage.
So the main advantage to looking for a VM-integrated storage system is that it allows your storage array to actually do what storage arrays do best, which is accelerating performance, copying data, moving data and sharing. And all of those things are really the compelling reasons that people buy storage arrays. So it's not one of those things where there's a specific advantage that is easy to speak to, except that not having a VM-aware storage array puts you at a disadvantage in terms of performance and capability.
And there's a whole class of arrays that go beyond VAAI and have much tighter integration into the virtualization management environment. And that is sort of the next wave where we're seeing companies come out with these arrays that automatically respond to provision requests from the virtualization platform -- they automatically tune their storage, they automatically configure it. And that to me is really exciting because we end up with much more advanced storage integration for our virtual machines. So I'd say it's important to have VM-aware storage in any virtualization environment, but to have sort of a second generation of VM enhanced storage is even more of a business driver.
Would you say there are any drawbacks to using VM-aware storage, or any instances where somebody would want to use traditional storage instead?
Foskett: Probably the biggest drawback is that if you use a storage platform that is designed with virtualization in mind, there's a whole spectrum of them, but some of them are designed only for virtualization, and you end up with an array that is really not useful for anything except a server virtualization environment. And that's fine, if all you're going to run are virtual machines on it. But some people are not running virtual machines on it; some people are running other workloads. And in that case you may want an array that is equally at home with conventional workloads as well as virtual machine workloads.
That's especially true with small businesses. Smaller organizations may say I just want one storage array; I'm not ready to go all virtual, so I really want something that works equally in both worlds. And then they look at sort of a unified storage platform -- something that has NAS [network-attached storage], maybe it has block storage as well, maybe it has array integration or VM integration. And that's sort of a different class of storage, and they may not get all that is promised by some of these super-enhanced storage arrays, but at least as long as they get something that is VM-aware, they get an array that they can use all around. Even VM-enhanced companies are starting to say, "Our system could be used for dedupe, or for scale-out workloads, or for other virtualization platforms." And so we may see that this drawback is less and less important in the coming year as arrays are made to work outside the VM space as well as inside.
This was first published in January 2013