Disk libraries and tape libraries: aiming for peaceful coexistence

Computer Technology Review, May, 2004 by Chuck Hollis

Tape libraries have been the primary technology deployed in traditional backup and restore environments. Until recently, they have offered an overwhelming cost advantage over disk. Customers that require the highest recovery service levels use disk-to-disk data replication, but this requires changing their traditional backup processes to implement the technology. Today, a new disk library category is evolving from these two traditional solutions. While not as fast as disk-to-disk replication, disk libraries integrate with existing processes and provide better service levels than tape.

Disk-based backup and recovery is gaining attention because of changes in both the business and the IT environment. As the volume of data organizations must backup and recover continues to expand exponentially, the window in which applications can be taken off line to do the backup stays the same or shrinks. Business managers want their data restored and their applications recovered more quickly than ever. For these reasons, the old tape-based backup and restore operations no longer meet today's business requirements. This paves the way for a new, disk-based paradigm for backup and recovery.

[FIGURE 1 OMITTED]

Some customers have avoided disk-to-disk backup and restore not only because disk storage hardware costs more than tape hardware, but also because they would need to change their backup processes and infrastructure. Disk libraries now offer businesses the opportunity to experience the benefits of disk-based backup at lower cost and with less effort than ever before. These libraries can be "dropped into" existing environments without upgrading backup software, retraining staff, or changing existing backup and business processes. Disk libraries also allow operations managers to provide faster, more reliable backup and restore for critical business applications--without the costs of changing their environment or processes.

However, disk libraries will not necessarily mean the end of tape. Tape's portability assures it a place within the information lifecycle--as the preferred media for long-term, off-site storage of data. To understand how the two technologies will work together, consider their respective advantages and disadvantages.

Tape vs. Disk in Traditional Data Centers

Tape has had two traditional recovery roles in most enterprises. One is for long-term, off-site storage of data that is unlikely to be accessed often (or ever again). The second role is to backup and restore applications, so that applications can be recovered if a system failure or disaster occurs.

As an off-site storage medium, tape is still the media of choice because it is the least expensive to purchase on a permegabyte basis and is the easiest to transport off site. Most medium to large organizations already have tape libraries, which can store and transfer data to and from hundreds or thousands of tape cartridges.

The use of new, lower-cost ATA drives is driving down the cost of disk-based backup and restoration of critical data. ATA is the enabling technology making disk-based backup and restore more affordable, because ATA drives cost roughly 50% less than the Fibre Channel drives used in more demanding, online transaction processing applications. Disk libraries consist of arrays of inexpensive, ATA disk drives linked by high-performance, redundant storage controllers. Because these same disk libraries emulate the software used to manage existing tape libraries, the backup application thinks it is still writing to tape, while actually writing to disk. This makes the migration from tape to disk libraries seamless.

The advantages of disk libraries include:

Higher sequential performance: Tape drives based on the Linear Tape-Open (LTO) 2 format transfer sequential data at approximately 25MB/sec in real-world environments. By comparison, a high-performance disk library may move as much as 400MB/sec, allowing a single disk library to offer the sequential performance of nearly 16 high-performance tape drives. In normal backup environments, high-performance disk libraries should improve backup performance roughly 30-60% over existing tape solutions.

Faster random-access performance: Disk libraries also shine when it comes to restore performance, which involves accessing data stored in many scattered locations on a storage media. Disk libraries deliver incremental data restores in one-tenth the time of a tape library because there is no need to load tape cartridges and rewind tape, in order to find the required data. Disk libraries leverage the random-access nature of disk compared with the sequential access nature of tape. Tape libraries require mounting the physical tapes during restore operations. In a disk library, all the volumes are online and immediately accessible.

Higher availability: Because disk libraries are built around storage controller technology, they can be fully redundant and use RAID protection against media failure.

Higher utilization of media: Tape cartridges come in fixed sizes, which do not always correspond to the amount of data an organization needs to backup at a specific time. As a result, companies must purchase, handle, and store many tape cartridges that are only partly used, which increases backup and restore costs. "Stacking" volumes on the same tape can increase utilization, but requires so much effort and overhead that many organizations simply choose to live with low media utilization. However, intelligent disk libraries allocate capacity as needed, allowing near-100% utilization of backup media. In addition, many disk libraries support data compression, which further increases utilization.