Relatively speaking, storage virtualization is a simple concept: take a bunch of networked storage devices and (magically?) turn them into one big storage device that can be managed from one central control tower. As a result, heterogeneous storage environments are easier to manage (at least theoretically), administrative tasks can be streamlined, data can be migrated quickly, and there is more efficient use of disk space, among other benefits.
These features help to explain its popularity. As with anything promising simplicity and lower capital expense, there is growing interest in storage virtualization. According to a survey commissioned earlier this year by F5 Networks, the percentage of U.S. enterprises implementing storage virtualization will jump from 21 percent to 47 percent in the next few years. European adoption will jump from 37 percent to 61 percent.
Nothing is as simple as it seems, though, as there are many paths to take from piles of disks to one big, happy pool. Storage virtualization can be implemented and discussed in several different ways — maybe three, maybe four, possibly five — depending on how you look at it and whom you ask. Some basics might help.
Broadly speaking, there are two types of storage virtualization: block and file. Block is most common, embodied by SAN or NAS technologies, where distributed systems look like one storage device. File virtualization takes place “in front of” the storage network, usually in an appliance of some type.
More often, you’ll hear talk of the implementation methods. Host-based virtualization refers to software on a server providing the virtualization layer and creating the one big drive effect. Appliance-based virtualization puts a hardware device on the storage network. Network-based virtualization works at the switching level.
Some people use other terms to describe the same scenarios. In-band virtualization means the virtualization occurs in the data path between the host server and the storage units, as with an appliance. Out-of-band means the virtualization is outside the data path, at the switching level. Split-path virtualization combines in- and out-of-band and typically involves software services or an adapter in a switch or appliance.
Some people reduce the discussion further to appliance-based virtualization versus network-based. Appliances are easy to use, proponents say. But, says the other side, they can add a bit of latency. The network approach can be more flexible and scalable, fans say, but others counter that it is more complicated to implement. It could come down to a matter of where you prefer to manage storage. It could come down to preferring a particular vendor and not caring at all about bands.
If you’ve got a favorite storage brand, chances are it has a virtualization solution. This includes both big server vendors — like IBM, HP and Sun — and storage leaders. EMC offers a range of block- and file-virtualization products, including Invista, which virtualizes SANs using a split-path approach, and virtualization consulting services. LSI’s StoreAge SVM (Storage Virtualization Manager) is a SAN appliance that provides virtual volume management in heterogeneous environments.
Last month, HP introduced the StorageWorks SAN Virtualization Services Platform (SVSP), a split-path, switch-based platform built on LSI’s SVM and that will operate across HP and non-HP storage arrays, the company says. Observers commented that SVSP will compete with IBM’s network-based SAN Volume Controller, which is designed to combine storage from multiple disk systems into one shared pool.
Hitachi Data Systems integrates virtualization into its enterprise disk arrays. One of the best-known developers of virtualization appliances is NetApp, whose V Series devices can scale up to the terabyte and petabyte ranges of storage using heterogeneous hardware. There also are tools from software developers. DataCore SANsymphony software turns physical disks into SAN-accessible virtual disks.
The reality, though, is that datacenters usually are filled with all kinds of storage brands, so the system capable of providing that single across-the-universe view and control will have an edge. In the F5 survey, 63 percent of U.S. respondents planning to implement storage virtualization said the ability to handle a heterogeneous environment is “important or very important.”
One company focused from the start on virtualizing storage across mixed systems is Bycast, which last month introduced the new version of its StorageGrid. Bycast’s mission is to store fixed content, very large files, for very long periods of time on disparate types of hardware, all while managing the environment from a single unified view. “We provide storage virtualization software for large-scale digital archives,” says CEO Moe Kermani. “We’re focused on storing, protecting and preserving data for long periods of time.” (“Long” means 35 to 100 years, he says, which is longer than the average hard disk is going to hang around. “You can build systems that last a long time when you virtualize the hardware,” he adds.)
“StorageGrid’s virtualization layer creates a unified fixed-content storage interface across multiple facilities and heterogeneous storage hardware,” Kermani says. “It presents a single system view to applications by virtualizing the underlying storage infrastructure, even if it is spread across multiple facilities, across multiple datacenters. And by building this on a grid architecture, we basically offer unlimited scalability.”
Bycast originally aimed its storage solution at medical imaging but now pitches it as a way to create archives of any fixed content, “whether it’s photos or Final Cut clips or oil and gas testing results,” Kermani says. “One thing about all the different systems in use today is they all store something. That unifies everything.”
IBM and HP both OEM Bycast’s technology. IBM sells it as IBM System Storage Multilevel Grid Access Manager Software, and HP as part of its Medical Archive Solution.
You could say Bycast’s approach got endorsement from EMC with its new Atmos software, which EMC says is for building “cloud storage.” Kermani says he sees how Atmos can be viewed as competitive. After all, EMC highlights the scalability of its platform and its ability to manage across global locations as if they were one big storage system — two things for which Bycast is known. But he points out that StorageGrid has been around for much longer, and that Bycast has been working on object storage, metadata-driven policies and issues of multitenancy “for six years now.”
Going against today’s marketing grain, Kermani does not mention the word “cloud” when describing StorageGrid. Instead, he talks instead about “one big storage system,” “hardware independence,” “storing files transparently across tiers, across applications,” and “digital archiving as a service.”
As the inexorable march toward virtualized everything continues, storage virtualization products and choices will keep growing. Because of the different architectural and topological approaches to storage virtualization, the variations in management tools, and the mutt-like natures of many organizations’ storage infrastructures, finding the right virtualization solution can be complicated. The easiest approach is to buy everything from one vendor — if you are a Company X shop full of Company X gear, buy Company X’s virtualization product. However, that doesn’t work for most companies, who will have to pick their poisons carefully. The potential rewards (and, increasingly, mandates) of spending less time and money on storage infrastructure and management make it an important decision.