matt_prigge
Contributing Editor

Are virtual tape libraries a passing fancy?

analysis
Mar 8, 20105 mins

With most backup software vendors supporting feature-rich deduplication, the need for VTLs has diminished

Not long ago VTLs (virtual tape libraries) were central to any enterprise disk-to-disk tape backup strategy. Highly scalable performance and the ability to deduplicate on the fly — usually without help from backup software — made them a great wedge to handle the D2D part of the D2D2T equation. But now that all of the major backup software vendors have built in highly effective deduplication, does the VTL appliance still have a home? Or is it our backup strategy that needs a round of deduplication?

To answer that question, it’s important to take a step back and really look at what the VTL does. Though the products available in the marketplace vary in terms of features, performance, and scalability, the general idea is that a VTL is a large chunk of high-performance disk with built-in software that fully emulates a physical tape library. They usually allow access via a combination of fiber channel or iSCSI interfaces and also provide some form of block-level deduplication. When using a VTL, the backup software in play doesn’t need to know that deduplication is taking place or even really to know that the backup device isn’t in fact a real tape library.

This emulation makes it possible to build a traditional tape-like backup strategy, while still taking advantage of the flexibility of disk-to-disk backups. What your backup software sees as tape media are actually deduplicated chunks of disk. That means that it’s generally possible for the VTL to maintain a truly massive number of backups as compared to your true tape resources, which can’t be effectively deduplicated. You can afford to keep months of backups available in your disk-to-disk layer and rarely, if ever, go back to real tapes to restore data you might need. All of this while still being able to manage the resources within your software as if they were real tapes.

So how is operating a VTL any different than using your backup software’s own disk-to-disk deduplication functionality pared with a bunch of disks attached to your backup server? The bottom line is that in small-scale deployments, it really isn’t. In small environments, you can gain the same benefits that VTLs offer without a separate piece of hardware — and often do so much more cheaply. In fact, implementing the disk-to-disk backup and deduplication in the software layer can bring some time-saving functionality to the table.

Some backup software can perform deduplication on the client side. While this places more load on the resource being backed up (which may be something you’re trying to avoid), it also disperses the load from the backup server and network. In this client-side deduplication scenario, only disk blocks that the backup server hasn’t stored for any resource at any point in its active backup history need to be sent across the wire to be stored on the backup host. This may make it possible to centralize backups for WAN-attached servers and get away with a lighter-weight backup server. Since VTLs are not an active part of the backup process — they just receive and store the backups — they can’t extend deduplication’s benefits to this level of the equation.

Nonetheless, in larger environments, the VTL appliance still has an edge in two scenarios.

The first is in performance. While it’s true that VTLs are generally constructed from fairly standard server hardware (most I’ve encountered run some form of heavily customized Linux OS), they are purpose-built to handle the rigors of deduplicating and storing massive amounts of data. While you can certainly do this using general-purpose server hardware and operating systems, at a certain scale there comes a point at which your backup server will start to look like it’s desperately trying to be a small SAN. That’s generally where the hardware scalability of a VTL may be valuable from a performance perspective.

The second is in the tape emulation functionality that VTLs provide. This emulation may seem like an unnecessary abstraction layer — after all, its goal is partly to conceal any unusual behavior that backup software wouldn’t understand. Now that the backup software most of us use can do the deduplication on its own (sometimes more effectively), is emulating a tape library still desirable? That depends upon what kinds of systems you’re backing up. If you have a lot of legacy systems (VMS, AIX, HPUX, etc.) with their own strictly tape-based backup mechanisms, sliding in a VTL can be an excellent way to inject the benefits of disk-to-disk backup and deduplication into architectures that still aren’t able to effectively leverage them on their own.

The bottom line is that VTLs are not dead by any means. Large environments that require massive amounts of backup storage and throughput will still be able to leverage benefits unavailable in software-only solutions for some time to come. Small environments, however, will be better suited to general-purpose storage coupled with industry-standard backup software that can dedup on its own.

So if you’re re-evaluating your backup strategy and trying to decide whether a VTL is right for you, make sure you pause and consider whether you actually need one. You may be surprised.

This story, “Are virtual tape libraries a passing fancy?,” was originally published at InfoWorld.com. Read more of Matt Prigge’s Information Overload blog and follow the latest developments in storage at InfoWorld.com.