By Yuval Shavit, Features Writer
Storage virtualization is still relatively early in the adoption curve, but it's gaining traction. With companies creating so much information, IT departments are coming up with various ways to consolidate storage and get the most out of each storage unit. Schemes like thin provisioning and tiered storage work well, but managing all those devices can soon become a headache. Storage virtualization provides a way to combine all those drives into one centrally manageable resource, which can benefit your customers via time and money savings. Better yet, storage virtualization lets you make changes to the storage configuration without downtime, making data migrations more palatable for your customers.
Virtualization in general is the separation of a device's functions from its physical elements. For instance, server virtualization separates computation and I/O from the plastic and metal server, thus letting administrators run several virtualized servers on a single physical server. Storage virtualization is similar in that a unit's physical drive is separated from its function, which is to store data. But while this concept applies to several storage concepts -- such as SANs, RAID and thin provisioning -- the term "storage virtualization" is generally used to refer to a process by which several physical disks appear to be a single unit. Virtualized storage is typically block-level rather than file-level, meaning that it looks like a normal physical drive to computers.
There are several topologies that virtualized storage systems can use (these will be described in the next installment of this Hot Spot Tutorial). Conceptually, though, the device has several disk drives feeding into it, and it in turn appears to be a drive itself. When an I/O request is sent to the virtual disk, it's redirected to the appropriate physical drive. That drive then makes the write or returns the data, which the virtualized drive then returns to the caller, just as if the virtual disk had been a physical disk.
The main benefit of storage virtualization is that it allows you to more easily manage heterogeneous storage environments, said Ray Lucchesi, president of Silverton Consulting in Broomfield, Colo. While it's relatively easy to manage identical drives, that task can get much more difficult if you have several vendors, or even several models from the same vendor. IT departments often buy from a few storage vendors, "just to keep them all honest," Lucchesi said.
For instance, many drives have advanced storage functions like thin provisioning or striping, Lucchesi said, but their management software can only handle drives from the drive's vendor -- and sometimes only if they're the same model. If your client has different kinds of drives, its IT staff will have to manage each separately, which can be difficult and time-consuming. Storage virtualization lets IT employees treat all of the drives as a single unit, which can then be striped or thin-provisioned more easily.
Depending on how you deploy the virtualization, the overall system may have slightly higher latency, Lucchesi said; the simplest method, host-based storage virtualization, is unsurprisingly the one with the poorest performance. But the main disadvantage to storage virtualization in most cases is that it costs money, he said. How much depends on the method.
Virtual disks also let you migrate data between drives more easily, and without any downtime. Let's say you had data on drive A but wanted to move it to drive B. Traditionally, you would take A down, transfer its data to B, and then bring B online. With storage virtualization, you can just specify that the logical unit number (LUN) that corresponded to drive A should now go to drive B instead. The virtual disk will begin copying drive A's data to drive B, but all new I/O requests will be redirected straight to drive B.
This can be very useful if your client has a tiered storage system. Many companies keep important, frequently accessed data stored on high-performance devices and data that's accessed less frequently on cheaper, slower devices. The problem is that certain data may be more or less urgent at different times. Instead of deciding between taking a performance hit from slow drives, over-provisioning fast drives or dealing with downtime several times a year, a company that uses virtual disks can migrate data to high-performance drives as needed.
Storage virtualization has specific benefits for value-added resellers (VARs) and storage consultants. Storage hardware -- like any computer hardware -- has a refresh cycle of a few years, but one of the pain points for customers is migrating their existing data to the new hardware during an upgrade. Some VARs therefore offer storage virtualization as a one-time service; they install a virtualized disk, use it to migrate the customer's data without downtime, and then remove the virtualization and leave the customer with the new hardware, fully migrated. The process of installing and removing the virtualized storage layer can result in some downtime, depending on the virtualization method used, but that downtime can often be significantly less than the amount of time it would take to copy all of the customer's data from the old drive to the new one, Lucchesi said.
There are four major approaches to implementing storage virtualization: host-based, network-based, appliance-based and subsystem-based. In the next installment of our Hot Spot Tutorial, we'll look at the pros and cons of each method, as well as how to safely plan and execute the virtualization deployment.
This was first published in August 2008