BUSINESS CONTINUITY DRIVES D2D BACKUP

It compliments, not replaces, tape

Computer Technology Review, June 2002 | By Mark Ferelli

It seems to be in the nature of technology to set conventional wisdom aside. Older methodologies give way to new in the constant race to make computing easier, quicker, and less work. This push for new techniques is very present in the backup/restore space. "The pain of backup is growing," said Jay Desai, at Network Appliance. "There is a need to scale as new primary storage becomes available."

That push is known as disk-to-disk backup, the practice of protecting data using hard disk as a primary backup medium rather than tape technology. Traditional approaches to backup consist of a backup application, a backup server, tape libraries, tape drives, and media. Emerging directions in backup solutions, however, integrate hard disk technology and systems into the backup function to serve as the primary target of the backup application, complementing tape drives and libraries that focus on the archive function. The performance advantage is fairly obvious: speed of backup and speed of restore of most recently backed-up data.

Nor are the technologies new, really. Hard disks, the serial ATA, and tape emulation have been IT tools for some time. Desai noted that they offered hardware-based snapshot technology in 1992. The product was designed for block-level mirroring for quick restore and incremental backup.

But a new push is on, based on new business and technology drivers.

Business Continuity a Driver
Diamond Lauffin at disk-to-disk vendor Nexsan points to "true business continuity" as the driver. In pointing out the importance of time, he reflected on a case from his salad days in the tape business. "The University of Oslo had 600TB to restore after its backup software triggered a crash. Using ADSM, the restore was completed in 20 hours."

At one time, and not long ago, restoring 600TB in 20 hours would have been considered heroic work and a job well done. But modernly, Lauffin noted that any delay is unjustified, given the available technology.

The increased interest in business continuity may be one driver, but so is economics, according to Kevin Daly at Quantum. "Tape costs, on an incremental basis, still delivers a lower cost/megabyte, but has high entry costs attached. Disk has the entry point advantages."

George Symonds at Legato Systems concurs with the economic driver point of view. He said: "It all comes back to recovery time. The price of disk is down, making it cost effective and reasonable to look at disk for primary backup." He also points to the popularity of snapshot technology for point-in-time incrementals.

Not that disk backup offers all the answers. Daly points out that the disk performance challenge is its limitation by the file system. Unprepared disk-to-disk backup could propagate a virus from primary disk to primary backup disk. Tape, on the other hand, does not have the same OS-based limitations. Many experts point out the weaknesses of exchange servers as regards virus attacks, so integrators must take extra care for implementing disk to disk solutions for exchange.

New Disk-to-Disk Initiative
In response to this trend, Quantum has formed the Enhanced Backup Solutions Initiative (EBSI), announced at the most recent Networld + Interop show. The EBSI is billed as a vendor-agnostic, industry coalition of storage companies dedicated to helping businesses better protect their mission-critical data with emerging hardware and software technologies for backup that maximize operational efficiency and cost savings. Founding members of the EBSI include Atempo, Legato Systems, Network Appliance, OTG Software, and QLogic.

Currently, EBSI members are working together to test, integrate, and certify their products to enable enhanced backup solutions. In the longer-term, they will also look to incorporate related features and functionality in their respective product roadmaps to deliver even more robust and easier to manage solutions. According to a printed statement, potential areas of further development include, but are not limited to:

Serverless archive- Makes it possible for data to flow directly from the backup target to archive without going through the backup server.
Synthetic full backup- Allows for incremental backups to be reconstituted into a synthetic full backup at a future point-in-time.
Distributed backup targets and centralized backup archives- Enables the distribution of disk-based backup targets throughout an enterprise and varying geographic locations while consolidating archives into a single manageable location.
Virtualization of tape systems- Extends the life of existing investments by reducing the need for archive-system upgrades at the same rate as backup target upgrades.
Backup data post- processing-Enables the solution to reformat or reorganize the backed up data so that it can then be restored faster when needed.

Well-known industry consultant Jon William Toigo is in hopes that this new initiative will bring some unanimity to the industry. He observed that the initiative "holds the promise for diffusing the acrimony and infighting between advocates of disk-based data backup solutions." He notes that the vigorous debates in this area "have confused consumers and pressed many into adopting technologies and strategies that deliver little or no protection, or protection at a cost that is misaligned with potential threats and their consequences."

Implementing Disk-To-Disk Backup
Nexsan's Lauffin points to three methodologies for hardware implementation of disk-to-disk backup: 1) primary backup on premises, 2) an investment in a tape-staging replacement for tape, and 3) as a cache.

Each implementation has a variety of advantages and drawbacks. Where disk-to-disk backup may face challenges is in the management space. Legato's Symonds observed: "Disk is not infinite; you'll want to recycle. But the question is how to manage the space?" The current toolset is manual.

But Symonds points out that snapshot technology is more of a management problem with all its differing applications. Said Symonds: "Management software for snapshot technology is poor. Different APIs for the various applications…including backup and recovery." At this point, the management approach of choice for disk-to-disk backup is tape emulation: treating disk-based data as if it was on tape. Symonds looks to tape emulation as the first step to more sophisticated disk backup. He expects disk-to-disk support in Legato's well known Networker software in the November time frame.

Implementation of disk-to-disk backup will certainly require the kind of sophisticated software that Symonds references. At software developer Atempo in northern California, Philippe Boyon identifies his Time Navigator software as an exemplar for such a requirement. The company has implemented its software for the DX30 product from Quantum (see sidebar), and supports the device as a standalone solution or as a cache to migrate data from disk to tape. It also may be configured for remote mirroring applications and offsite vaulting, consistent with the importance of disaster recovery.

Tape Not Replaced
Although disk-to-disk backup bids fair to replace tape as a primary backup medium, integrators and IT management alike should not lose sight of the fact that tape technology is the workhorse of data protection and business continuity. It should continue as the price performance leader for archival applications and offsite vaulting.

Industry consultant Toigo puts it well: "Between tape only backups and disk-to-disk mirrors are a range of alternative solutions available at differing price points to consumers that can safeguard mission-critical data from loss, whether due to application errors or terrorist attacks."

The sentiment can also be detected in a comment from veteran industry analyst Fara Yale at Gartner: "[Disk-to-disk] products are positioned as complementary to tape products, not as a replacement."

Disk-to-disk backup is probably not for everybody. It certainly isn't for every application. But it is an increasingly realistic alternative where fast data restore makes a difference.

Download PDF