HDD And The Changing Face Of Backup.

FERELLI ON ... HARD DISK STORAGE

Regular readers of CTR are aware of our editorial stance that, in information technology, applications are king and will not condescend to any lesser role. Storage exists to serve applications in the enterprise, the midrange, the SOHO environment, and the growing world of electronic commerce. This latter application may well turn some of the industry's conventional wisdom regarding operations. The old rule was hard disks for primary storage and tape for backup. Recently, this issue has come up ... will tape remain the primary backup medium?

Analysts at GartnerGroup and Dataquest seem pessimistic. They suggested, at their recent StorageTrack 2000 conference, that hard disk will become the primary backup medium within five years. Their justification is founded on what seems to be a new imperative in the use of storage technologies. In the constant rivalry between capacity and speed, speed is beginning to pull ahead. The ability to handle increased transactions per second seems to have become more important than capacities per individual spindle.

Part of the reason for this phenomenon is the nature of electronic commerce. The demand in e-commerce is for speed . . . in order entry, in information location and retrieval, and in meeting requests for quotation. Speed is the indicator. Data packets are growing at a significant rate.

Now, what does this have to do with backup? Backup is meaningless without restore. Historically, tape backup has been a necessary evil and, if it took some extra time to back up, that is the price of data protection, but if the need for data protection has been eclipsed by demands for data availability, then the Gartner analysis must be correct. No one argues that tape access latency is slow. That slowness has been the excuse for the tape-is-dead bashing that you can hear every year or so.

Yet now, the bashing is tempering to a logical point the value of information. In a recent discussion with CTR, EMC's Don Swatik pointed out that, as the value of information increases, the time it takes to get the information becomes valuable. He suggests that an outage means lost customers, lost business, and tape is just too slow.

Does this mean that I, too, am suggesting that tape is fated to fade away? Absolutely not and anyone who thinks so hasn't done any homework. In the first place, if all information must go under a single infrastructure (as is the enterprise way), then tape must be included, de minimis for legacy reasons. There are billions of tape reels and cassettes out there and any integrator who suggests that an enterprise obsolete its tape investment is begging to lose clients. Although the tape market has been flat in the last couple of years, more innovation and new formats have been introduced to the industry than ever before. Speeds are increasing; reliability is increasing; automation of tape is a very healthy activity.

All that being said, the nature of tape's use in the data center might change. It is not inconceivable that the first backup copy of a file will reside on a disk array. Disk mirroring tools are accomplishing such things right now and so-called "snapshot" tools are following right along, but disk drives are anything but invulnerable. Disk failures are certainly not unknown and data availability is often a function of RAID level and spindle speed. Individual storage servers and NAS devices are likely to safeguard that data availability. So the question devolves to price.

Tape used to have the advantage, but hard disk prices are getting to a penny per megabyte. While tape drive prices are falling annually, it would be hard pressed to match disk pricing. Archiving is likely to remain the inviolate domain of tape, but what of disk versus tape in the backup space? Watch for next month's column.