The same storage options exist for different types of virtual environments, whether it be for server virtualization, virtual desktop infrastructure (VDI) or the cloud. But some data storage strategies work better for certain environments than others. For instance, according to Howard Marks, founder of DeepStorage.net, flash storage might be most beneficial in a VDI environment, while cloud services are best for backup and recovery. In this TechTalk interview from the Storage Decisions seminar, Marks discusses which types of data storage strategies and features work best in which virtual environments. Watch the video or take a look at the transcript below to find out.
Can administrators expect to find the same data protection features in virtual server storage solutions that are available in traditional physical server storage solutions?
Howard Marks: Well, the truth is we're talking about the same arrays with the exception of Tintri, which makes an array specifically for virtualization. The storage you use for virtualization is the same storage you use in the physical world. The problem is that the data is very different. So, data protection features like snapshots can't be used the same way.
In the physical world, a volume holds a single workload. It might be your SQL [Server] or your Exchange Server. And you can coordinate between the software and the snapshot provider to create an application-consistent snapshot.
In the virtual world, one volume on the storage system holds several virtual machines (VMs). There just isn't enough time to quiesce all of those VMs so that when you take a snapshot of that volume, it's application-consistent for all of them. So, array snapshots become much less valuable in the virtual world because they're going to be crash-consistent, not application-consistent. And we all know "crash-consistent" is a euphemism for "as good as if the server just crashed," which isn't very good at all.
What are your best storage practices for VDI environments?
Marks: VDI presents a unique set of workloads to the storage system. When we talk about VDI, we have to be pretty clear that there are two different VDI work cases. There are non-persistent desktops, the kind of thing you would do for a computer lab in a college or for the ticket counter in an airport, where the users run a small number of applications or there just isn't the personal relationship between a personal computer and the user.
In the non-persistent world, where we're talking about task workers, speed is all that matters. The data has very little value because it's just the same image repeated over and over again. So there, you want to have some sort of flash, maybe server-side flash that you're just using as a disk; and if the server crashes, you can just move those users to another server and a fresh image, and it's not a problem.
The other VDI use case is for, not task workers but knowledge workers, the people who install their own applications, or at least install the WebEx plug-in to their browser when they're going to connect to WebEx. If you give them non-persistent desktops, they will get pitchforks and torches and chase you down the street.
So, if you're going to build persistent desktops, the best way to do persistent desktops is not through linked clones, but to get a storage system that does inline data deduplication. Then build full clones on top of that inline deduplication system. Have the data-reduction problem done by the storage, not by the VDI environment. That means you can give people the same rich environment they had when they had desktops at a reasonable cost, because while flash is expensive, you've deduplicated your data and need a lot less flash to provide the performance.
What are your top three storage management tips for virtual desktop environments?
Marks: My top tip is the one we just talked about. It's don't use linked clones for persistent desktops. Instead, use a deduplicating flash-based data storage system. Pass the workload off to the storage system that's prepared to do it.
The second tip is not really a storage tip, but it affects how well your storage works. It is to invest in persona management. You want to separate the user's persona from the desktop image, and store the persona in a persistent storage system.
The third is, in the non-persistent desktop environment, to recognize that the data isn't important, so that you can use cost-effective solutions like single-controller external storage systems or server-side flash, that provide performance even if they don't provide a degree of resiliency you would insist on for your SQL Server, because you can just replace that image data quickly.
Where do cloud services fit in a virtual server storage management strategy?
Marks: Primarily, cloud services fit in the backup and disaster recovery environment. If you are not a Fortune 500 company, it's very difficult to convince the [chief information officer] to pay for a disaster-recovery site, and that's going to cost $5,000 to $20,000 a month to rent the space in the call center and put the servers there to sit in case you're going to need them later.
Most of the cloud service providers and the virtualization backup and data recovery systems have made alliances. So, if you run VMware, you can use Veeam backup and replication or Zerto, and they have partners who run public clouds. You'll replicate your data to the public cloud provider. In the event that your data center is unavailable, you can spin up all of your workloads in the public cloud provider. All you have to pay for is the storage for your backup data until you spin those workloads up. It's a lot less expensive solution, and makes a huge amount of sense for all but the largest customers.