The key word here is "design" - as opposed to "throw together" or to pick a "canned" system from an overly simplistic menu. The intense primordial urge of the techie (presumably since prehistoric times) to skip all of that planning "nonsense" and just roll up his (or her!) shirt sleeves (or in the old days, pelt sleeves) and start implementing something (or in non-Techie English, doing something) must be resisted. And no, resistance is definitely NOT futile!

Right now, the topic is backing up your system. Realistically, we’re talking about tape as the media of choice for the vast majority of systems. The cost per megabyte stored is absolutely unbeatable, and its reliability is proven. Even for single-user systems, using removable-disks, such as Iomega’s Zip drives, isn’t as cost effective. I’ve found that business people are incredibly cost sensitive about tape cartridges. At anywhere from $20 to $50 per cartridge, having enough cartridges to provide a thorough backup sequence can cost more than the actual tape drive! Using any media with a higher cost per megabyte would encourage a less than adequate backup pattern. Note that there are special cases where high-capacity magneto-optical disk cartridges or other technologies may be practical, but those are rare exceptions.

I recommend that, as a minimum, my clients use at least a three-week rotating backup pattern that uses 15 tapes for those companies that don’t do weekend backups. An additional 3 (or more) tapes can also be used as a separate "fiscal month" (consisting of 3 weeks) rotating backup. If one of the weekly tape groups (5 tapes) is kept offsite, the business can conceivably survive a building-destroying disaster.

Each of these tapes contains a complete backup of the system. This implies that you should have tape drive with the capacity to backup the whole system at once. If necessary, you can use either multiple drives or an autochanger (a jukebox for tapes) so that one nightly backup actually spans across more than one tape. I do NOT recommend manual spanning for any but the smallest single-user system, as the manual effort required to change tapes usually results in the backup process being skipped when it’s not convenient.

An absolute requirement is to manually check the backup results report (usually automatically printed if the backup software is set up appropriately) every day to make sure nothing unusual happened. I tell my clients to immediately fax me that report if there is an unidentified anomaly within it. If the backup is suspect, I want to know about it right away. Blindly running backup jobs and changing tapes as scheduled is a great way to set yourself up for trouble the next time you need to retrieve data from your tapes. I see this done a lot, but, usually don’t find out until after a disaster that requires retrieving data from tapes.

As to the tape drive, itself, I strongly encourage my clients to use an external drive. If the computer has any sort of internal electrical or heat related problems, an internal tape drive could be toasted along with the computer. Then, not only would you have to worry about repairing or replacing the computer, you would also have to scramble to repair or replace the tape drive. Since tape drive repairs usually take some time, you might end up buying a second drive. Also, if you’ve had a tape drive for a while, you might not be able to easily find a compatible replacement. This is after all, the computer industry, where a "generation" of products can be defined in months if not weeks.

For larger capacities, I sell DLT (Digital Linear Tape) drives in 30, 40 and 70GB capacities. DLT autochangers are available, too. These drives are noticeably more expensive than DDS drives, so I don’t sell many, even though the cost per megabyte of the tapes is less.

An up and coming tape technology is AIT (Advanced Intelligent Tape) which uses a new AME (Advanced Metal Evaporative) tape material with MIC (Memory in Cassette) technology that stores the tape directory on a chip withing the tape cartridge. I haven’t used it yet, but I have heard very nice things about these drives. The come in 50 and 100GB capacities (including compression) and autochangers from 4 to 40 cartridges are available. That’s up to 2TB (terabytes) before compression for a 40-cartridge library (the name used for a really big autochanger) with one or more 50GB drives if you’re keeping score! Soon, that library will be available with 100GB drives in case you have some really serious offline storage needs.

One technology that hasn’t worked well on servers despite a good track record on individual computers is Travan. There’s something in the interface that keeps Travan tape drives from keeping up with the new faster servers and more efficient Network Operating Systems. They tend to get out of sync in communicating with the server, essentially freezing the drive. At minimum, it takes a server reboot to revive the drive. Often, it requires multiple, obscure, commands to regain communications with the tape drive, a task well beyond most users.

Another thing to avoid is bizarre proprietary technology. I’ve lost track of how many new, innovative tape technologies, supported by only a single firm, have come and gone. Does anyone remember the Pereos drive from Datasonix? It was a drive smaller than my hand that used cartridges based on the micro cassettes used in handheld recorders. It stored an amazing 1.25GB compressed and cost only $695! Today, a Pereos drive would make a unique paperweight. With a pair you’ve got bookends!