Tape: Is The Glass Half-Empty?
For even faster access, mirroring the hard drives may be an even better, though more expensive, method of data storage. With the increasing acceptance of Fibre Channel, mirroring over large distances, even several miles, is possible. The increasing acceptance of Gigabit Ethernet over Fibre and the impending arrival of ten-Gigabit Ethernet make the transfer of large amounts of data between servers and the mirroring of large SAN systems possible.
Certainly, 100GB hard drives will be more expensive than 100GB tape cartridges. Two years from now, the cost of a pair of 36GB or 72GB hard drives or maybe even a single 100GB hard drive will be much closer to that of a tape cartridge. Although the hard drives won't be as robust or shockproof as a tape cartridge, hard drives will continue to offer various advantages over tape.
The first advantage will probably continue to be a lower level of obsolescence. If you're archiving data, are you sure that, twenty years from now, you will be able to find a drive that can read the tape you write tomorrow? Although hard drive interfaces will continue to go through evolutionary changes, the likelihood that you can find a computer with enough backwards compatibility to read your old SCSI or Fibre Channel drive will probably be much greater than the likelihood of finding a 20-year-old tape drive that can read an old tape. It should be clear that, if interface issues become only a minor inconvenience, the likelihood of being able to read data is much better if the media and the device that can read and write it are stored together, as is the case with a hard disk drive.
A second, obvious advantage is the high performance and random access capabilities of the hard drive. Ignoring what might happen to a tape that is stored for 20 years, it ought to be clear that a tape might have to be indexed (a process that could take a few hours) and access to specific files on the tape may take minutes. For a hard drive, such access can happen in seconds--or faster.
End Of Life Technologies?
Although it's been easy to predict the end of life for both tape and hard disk drives, these predictions have so far been misproven in some cases. Hard drives are approaching the superparamagnetic limit, a phenomenon that relates to the size of the magnetic particles on the disc platter being so small that they spontaneously rearrange themselves, making it impossible to store data at unlimited densities. Last year, this limit was redescribed as the "superparamagnetic effect," implying that the limit may not be quite as limiting as originally thought.
Helical scan technologies used for 4mm and 8mm storage seem to be heading towards finite limits. I had doubts about DDS drives going beyond DDS4. Hewlett-Packard has made it clear that there's at least one more drive generation left. However, it is easy to conclude that, without some amazing new engineering, DDS5 may be the end of the line for DDS.
Sony's AIT drives have a roadmap for a few more generations. These 8mm drives will eventually use GMR (Giant Magneto-Resistive) technology to increase the data density on MT cartridges.
Ecrix's VX1 drive implements a packet writing method that improves reliability and recoverability of data recorded onto these 8mm cartridges. The company also has a roadmap for future generations. Similarly, Exabyte with its Mammoth family also has plans for more Mammoth generations.
A key feature of SDLT and Ultrium is a multiple generation roadmap. Both will be available from multiple vendors: SDLT will initially be available from Quantum and Tandberg (which has licensed the technology from Quantum). IBM, Hewlett-Packard, and Seagate will offer Ultrium. Work is already in progress to develop Ultrium 2 drives.
Roadmaps that extend beyond three or four years include a mix of real engineering and wishful thinking. The wishful thinking in some cases hopes that new breakthroughs enabling the capacity and performance in the roadmap come to pass. The engineering extrapolates current capabilities to even better ones in the future. Earlier roadmaps for tape technologies have not been tremendously overoptimistic. There's no telling how far the technologies can continue to be pushed.
Will The Need Go Away?
The old paradigm of backing up data to tape and storing it in a safe place may be becoming obsolete. In the days when the data only resided on a central system, backing the data off of the system became critical to business survival.
Today, a user's notebook can be backed up over the Internet, its data safely encrypted when it is moved out of the notebook computer and stored in encrypted form on a storage device who-knows (or cares) where. The automatic backup can be initiated whenever the notebook is connected to the Internet--automatically and transparently to the user. If the notebook breaks or is stolen, the entire drive image(s) can be restored onto a replacement unit, simply by connecting to the Internet. In the future, the lost data can be restored onto bootable CD or DVD discs or copied onto a hard disk.
The concept of synchronizing data between multiple machines, using the Internet as the conduit, is becoming increasingly familiar to many companies and many users. If, for example, the data on my home desktop is synchronized with the data on my office desktop and both are synchronized with my notebook computer, will it be much of a catastrophe if I drop my notebook and make it unusable? If the data are synchronized to a drive in a backup server farm somewhere in Kansas, does it matter if a power surge, when the neighbor's microwave blows up, fries my office desktop's hard drive?
The point here is that, for many people and companies, multiple devices will be synchronized, providing near constant backup on multiple machines. With active backup, synchronized whenever the device is connected to the Internet or an office network, will the need for tape backups continue to grow?
After covering the tape industry for the last 12 years, I can't go along with the proposition that tape is a dying technology. If we were in the late 1980s, such an attitude would have been understandable. These were the days that tape had to be mounted manually and the problems of human error made tape management a nightmare, but 1987 came around with the beginnings of tape storage automation and tape surged as the cost effective, more easily managed technology of choice.
The measures to use, I suggest, are data accessibility and economic constraints. As far as access is concerned, tape automation especially offers high speed access to data; improvements in tape robotic performance speed tapes to drives and tape library software enables better backup control than ever before. At the same time, tape capacities are reaching new highs and the virtual tape concept vastly improves tape usage.
If tape were dying, no startup in its right mind would introduce new recording formats. Instead, VXA from Ecrix and formats from Benchmark and Chaparral are reaching the marketplace. This year, albeit late, LTO products and SDLT should ship in quantity. Yes, they are later to market than announced. So was the original introduction of Exabyte's Mammoth drive, which is now shipping in popular quantity and is justifying the company's faith in the format.
The cost per megabyte of automated tape storage purchased typically falls well below $.01 today. Technology improvements, like Overland Data's VR2 chip, are likely to drive prices even lower as each tape cartridge is more effectively used.
For archiving applications, tape is currently supreme. The only thing more popular is paper, but since government and academic sources are projecting capacity requirements in petabyates, where will 500 billion sheets of paper be stored? No, the report of tape's death is still being exaggerated. The glass is filling up.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Technology Information|
|Publication:||Computer Technology Review|
|Date:||Apr 1, 2000|
|Previous Article:||The Case For Optical NAS Solutions.|
|Next Article:||Helical Scan Vs. Linear Recording Tape Drives.|