BY HEIDI BIGGAR
Talk of non-tape-based backup has picked up significantly over the last few months. Vendors such as Legato, Maxtor, and Sony, as well as newcomers Alacritus and Nexsan, have all announced some form of disk-based backup capability, and others are expected to follow suit.
Computer Associates, for example, will enable disk-based backup in the next release of ARCserve (it is already possible with CA's BrightStor Enterprise Backup), while other software vendors such as Legato say they are working on significant enhancements to their current disk-based backup capabilities.
"Going disk-to-disk is easy, but when you start to scale, that's when it gets challenging," says George Symonds, vice president of product management and development at Legato. In fact, most backup vendors have been able to do disk-to-disk backup for some time but have been waiting for disk prices-particularly ATA disk arrays-to drop low enough before offering it.
"We've been waiting for vendors to introduce low-cost hardware [e.g., ATA RAID] that can be used for this type of disk function," says Marco Coulter, divisional vice president, BrightStor, Com puter Associates.
The ability to back up data to a disk array rather than to a tape library is simply a matter of adding disk to the software's list of backup targets, adds Coulter. "It's reasonably straightforward...just GUI stuff."
Users specify disk, rather than tape, as the backup destination. Data is then written to disk as a tape image (i.e., in tape format) and then recovered from disk through the backup application's browser just as if it were written to tape. The process of performing a backup, taking the image to disk, and then over time, moving it to tape, is called "staging."
Whether disk or tape should be used depends on a number of factors, most notably the types of underlying data structures (e.g., structured, semi-structured, or unstructured).
"Customers are using [low-cost] disk for a 'staging' area," according to Coulter. "They create the backup there but then use our tape copy function to copy it from disk to tape later on." In doing so, users not only benefit from faster backups but also from potentially faster restores, he says. "But now, they are only using it for faster backup."
Carolyn DiCenzo, Gartner Group/Data quest's chief analyst for storage management and storage area network (SAN) appliances, agrees: "I don't see much interest in backing up to disk now, but maybe in the future. I believe the big markets will be backup and recovery from replicas."
All the leading backup applications (e.g., CA BrightStor Enterprise Backup, Veritas NetBackup/Backup Exec, Legato NetWorker, etc.), as well as many smaller players such as Atempo (formerly Quad ratec) are able to back up tape images to disk. The real challenge, vendors agree, will be backing up, managing, and then recovering replicas to/from disk (see sidebar).
So, while the concept of disk-based backup may not be new, its role certainly is. Recently, vendors have begun positioning ATA RAID and even network-attached storage (NAS) devices as backup targets in traditional tape backup markets (see "Sony enters NAS market" and "Maxtor sets sights on 'enterprise' NAS," InfoStor, October 2001, p. 8).
Plummeting disk prices have opened the door. With the aggregate average selling price of desktop ATA drives expected to drop to $0.003 per megabyte, according to International Data Corp., the cost of ATA RAID arrays is approaching that of tape libraries, according to Dave Reinsel, senior research analyst, hard disk drives and components, IDC.
At these levels, ATA disk becomes a viable backup alternative to tape-based products, explains Don Trimmer, president and chief strategic officer at Alacritus, one of two recent start-ups that focus on the ATA RAID backup market.
Alacritus, together with Hitachi CP and Nissho Electronics, has begun shipping a "virtual tape library appliance" into traditional tape backup-and-recovery markets, while Nexsan-the other start-up-is selling an ATA RAID array as a primary on-site and secondary off-site backup device (see "Alacritus enables disk-based backup," InfoStor, September 2001, p. 14).
"These devices create another level in the storage hierarchy," explains David Hill, research director at the Aberdeen Group consulting firm, in Boston. "The random access of the devices may be slower [than higher-priced arrays], but you can still run applications off them. You can't do this with tape." Users, he says, also benefit not only from faster restores, but also from easier management and being able to access individual files versus whole volumes.
Nexsan claims its ATAboy array, running Disk-to-Disk (D2D) backup software, has virtually all the high-availability and fault-tolerance features of midrange arrays, but at a fraction of the cost and without the management issues of a Veritas or Legato approach. Unlike Alacritus' Disk-based Data Protection (DDP) software, which works with leading backup applications, Nexsan's D2D operates as a standalone backup utility.
Nexsan prices its 1.2TB ATAboy array at $0.017 per megabyte, according to Diamond Lauffin, senior executive vice president at Nexsan. The company is targeting a variety of vertical markets, including digital-video security, medical imaging, print graphics, animation and special effects, and telecommunications.
Alacritus, meanwhile, says its virtual tape approach to disk-based backup (the backup disk appears to the host as a tape library) addresses user demand for both faster and cheaper backup. "The problem with the leading backup vendors' approach to disk-based backup is that it only works for users who have small and inactive data sets," Alacritus' Trimmer contends. "It just isn't designed for huge files."
Trimmer gives the example of backing up an NT file system (at 1KB block sizes) to an LTO drive that writes data at 20MBps. "That's 20,000 I/Os per second," he explains. "Not only does it take a pretty high-end disk array to do 20,000 I/Os per second, but you'd be completely consuming it. And if you've got multiple data streams going, you're looking at multiple arrays."
Similarly, it takes a fairly high-end server and multiple high-end disk arrays to get the scalability and performance to replace a $10,000 tape library, he adds.
Besides cost and performance advantages, Trimmer claims that Alacritus' approach can also simplify management in backup environments running multiple vendors' disk-based backup applications and can even add a layer of data protection in single-application environments.
Alacritus' DDP software is designed to work with, not replace, backup applications from leading software vendors. In addition to its virtual capabilities, the software provides a unified way of doing library sharing, dynamic disk-drive sharing, and performance tuning-key areas where vendors try to differentiate themselves, explains Trimmer. "So, to a small extent, we also tend to break their lock-ins." Trimmer considers library vendors, not software vendors, as the company's primary competitors.
Putting it all together
Does disk-based backup spell the end of tape? "Absolutely not," says David Hill, research director, storage and storage management, at the Aberdeen Group consulting firm, in Boston. "Contrary to popular belief, the storage pyramid will reflect a larger [rather than lesser] role for tape," says Hill in a recent white paper on the subject. "The well-known storage pyramid is not about to be compressed accordion-like to just one level-i.e., disk storage."
Hill says that although more and more backups will be done via point-in-time snapshots to disk, the total number of backups will actually increase, not decrease, and that the increase will not be to the detriment of tape.
Similarly, Hill believes that replication technologies, which do not protect against data corruption, will not replace tape. Instead, he envisions a backup world in which disk tape are used together for near-instantaneous data restore.
Tape will continue to play a key role in backing up mission-critical databases and unstructured and semi-structured data [see chart], says Hill.
"I think what we are going to end up with," says George Symonds, vice president of product management and development at Legato, "is a multi-tiered, or 'HSM'-type, backup paradigm in which snapshot-because of its speed and point-in-time capabilities-acts as the primary recovery mechanism; replication, the secondary; and tape, the tertiary." According to Symonds, "Whether tape is considered archival storage or backup depends on how quickly the data is moved from primary or secondary disk storage to tape." Either way, he argues, tape plays a central role.
Will snapshot and replication technologies filter down into midrange and low-end products? Absolutely, says Symonds. "You're going to see replication implemented in many places-in storage hardware and on the server side, in the high-volume Windows area." Snapshot capabilities, he says, are already being implemented in midrange disk arrays, and some vendors such as St. Bernard Software have been selling low-cost software-based snapshot technologies into the server market for years.
But a key problem with point-in-time copy is its inability to protect data against physical hardware failures. To get this level of disaster-recovery protection, users are forced to make physical copies of their data, which can magnify their already-expensive disk costs, explains Hill. This is where ATA RAID fits in.