Tape vs. disk: another view.

Tape versus disk. The debate over deploying tape or disk for data storage has existed since the inception of these media types. As rich media and other data swell to immense proportions for enterprises and governments, the storage media controversy seems to intensify. However, the selection is not that simple. In fact, there is no one choice. Instead for most large data stores, a media blend (i.e., disk, tape, etc.) rather than a single media type better satisfies customer prerequisites for cost-effective and efficient data storage and access.

In the November and December 2001 issues of Computer Technology Review, Diamond Lauffin of Nexsan Technologies authored a series of well-constructed and informative articles titled, "The Arrival of Affordable Disk-To-Disk (D2D) Storage." Mr. Lauffin's thesis maintained that the use of tape media for primary backup makes less sense today because of the lower total cost of ownership of newer, faster disk arrays. He is correct. Technological advances and manufacturing efficiences have made disk media better than ever, giving both consumers and enterprises more "bang for the buck." And for primary backup that aids business continuation in case of a disaster or other type of problem, disk makes sense.

But the author's enthusiastic approach to disk seems to have lessened the value of current tape technology and the software that manages it. Mr. Lauffin writes, "Maintaining tape can involve such tasks as identifying and retrieving tape volumes, monitoring tape barcode sequences, loading and unloading tape, inspecting tape for wear, rotating tape, manual disaster recovery operations, and inventory management." Had this been written five years ago, the author would have been accurate. But looking at tape storage today as simply an off-line, archaic archive technology can be deceptive. In fact "it's interesting to note that despite some predictions to the contrary, tape storage remains typically the ultimate resting spot for data." ("Tape Still Alive and Kicking," by Marion Apicella and Mark Jones, Infoworld.com-Storage Report, December 18, 2001).

In addition, missing from Mr. Lauffin's series is that tape, in the form of "near-line" database extensions for infrequently used or "dormant data" is still very much an applicable technology, on par with and more cost-effective than disk (when used with disk arrays in a tiered storage system). So let's take a look at where tape does fit in the storage hierarchy, and where advances in tape technology and the software that manages it, make tape as relevant and as cost effective as ever.

According to Bill Inmon noted data warehousing consultant and Partner of BillInmon.com, "... as data warehouses pass the three to four terabyte range they need to be split over multiple hierarchies of storage media." ("Storage Rant," by W.H. Inmon, BillInmon.com LLC, Newsletter, Volume 3, Issue 1, January 4, 2002). The reason Inmon gives is that data should be divided between "active" and "inactive." Active data is frequently accessed, business-critical information needed for operational functions. Inactive data, though also vitally important, is seldom accessed and tends to be more of historical value for analytical tasks and customer relationship management (CRM) applications. Inmon states, "Keeping huge amounts of dormant [inactive] data on disk storage is like clogging the arteries with cholesterol. By moving unused data off of high performance disk storage, the system is scrubbed and only actively used data remains, In doing so, performance really zips." ("Storage Rant," by W.H. Inmon, BillInmon.com LLC, Newsletter, Volume 3, Issue 1, January 4, 2002).

For example, to support their CRM efforts, a major telecommunications company in the southern U.S. wanted to provide their customer service representatives with almost six years' worth of billing records.

After analysis, they calculated that this requirement would eventually yield between a nine and 11 terabyte data warehouse. Subsequent examination by the company resulted in an important revelation: Billing inquiries by customers fell to less than 16% of all queries after one year and then less than 8% past 15 months (See Table).

[GRAPHICS OMITTED]

Realizing that it did not make performance or economic sense to retain data older than 12 months on disk, the company opted to migrate information based on age to an alternative, near-line tape storage system. With this strategy, the advantages were multifold:

* The company's main database application (in this case, IBM DB2) was relieved of a heavy burden. It is no longer responsible for managing multi-terabytes of archive data within its disk system.

* DB2 has access to all information because it can transparently access the inactive, historical data on the near-line system.

* The combined system is more highly automated because it eliminates the need to manually mount tape cartridges.

Even though a disk-to-disk system enables recovery from a hardware or site failure, periodic backups of archive (unalterable) data to tape by the near-line system protect against accidental deletes, software errors, and viruses that can corrupt primary data managed by disk. If these types of problems should ever occur, data can always be restored from previous backups.

Moving inactive data to alternative, near-line media is just part of the story. What about tape performance, capacity, cost savings, and the applicable software interfaces? Here is where the disk versus tape debate gets interesting!

Since the early '90s, tape has kept pace with disk in reducing the dollar cost per gigabyte (GB) while improving access speed. At their inception, IBM cartridges could store less than one GB. Today, LTO and S-DLT can store 100GB in the same space. And with the advent of Sony's Super-AIT (Advanced Intelligent Tape) next year, uncompressed capacity will rise to 500GB with a transfer rate of 30MB/sec. Due in 2003, Tandberg's first-generation O-Mass offering will have an uncompressed storage capacity of 600GB, with succeeding releases rising up to an amazing 10 terabytes (TB) on a single cartridge. Transfer rates on O-Mass' first generation cartridge is expected to be 64MB/sec, accessing data in less than 3.5 seconds. ("Tape Still Alive and Kicking," by Marion Apicella and Mark Jones, Infoworld.com-Storage Report, December 18, 2001).

Using the 80-20 rule (80% of data access is targeted to only 20% of the data within a repository), a hierarchical storage management system with a blend of media types can provide an economical system for storing all enterprise information by keeping the active 20% of data on disk and the remaining inactive 80% of data only on tape. With a tape system at approximately 1/10 the cost of an IDE disk system, this approach costs 28% of an all disk system (20+80x1/10) /100. Of course this 80-20 rule can vary by application (for example, 70-30 or even 90-10).

Tape hardware cost and performance aside, there is still the near-line software management issue. In an alternative storage system, Inmon states "there are two places where software is needed--for the management of the data on the alternate storage and for the management of the movement of data to and from disk storage and alternate storage." This software, which Inmon refers to as a cross media storage manager (CMSM), is also responsible for providing what he calls the most important function--query management.

Some present-day CMSM software provides the ability to query data held on tape at the row-, or detail-level, thus eliminating the need to move large files up the hierarchy. Performance improves because only the required information is transparently presented to the end-user. Other features of today's CMSM software (e.g., FileTek's StorHouse, OTG's DXDB) clearly answer some of the reservations Mr. Lauffin has about tape maintenance. For example, with current near-line management software, there is no need to manually load and unload tape; monitoring tape barcode sequences is an automated process; and even inspecting tape for wear is automated with alerts sent to a remote site that predicts media failure before it happens.

While D2D storage is extremely applicable for primary backup, especially with disk media prices failing, there are still areas (such as large data repositories) where tape is relevant, more cost effective, and technologically strong. In fact, the marketplace agrees. According to Freeman Reports, tape drive sales will expand by a 4% annual growth rate, climbing to 69,700 units by 2006 with a value of approximately $1.9 billion. ("High-Performance Tape Market Will Grow," Freeman Reports Press Release, September 18, 2001). As manufacturers and software developers commit resources to advance tape technology further, tape will remain a viable alternative to disk-only storage.

www.filetek.com

John G. Burgess is the CTO and vice president of FileTek, Inc. (Rockville, MD).