Home Page



Thoughts on Backup
By Thiravudh Khoman

First of all, a confession: I've always done a pretty poor job when it comes to "proper" data backup. I've tended to rely on "spur of the moment" backups, my repair skills (and the skills of others), and a misplaced faith that bad things won't happen to me to keep data disaster at bay. Not that I always get off scott free - I don't - but so far, I haven't lost anything earth shattering and don't experience data death very often (knock wood).

Floppies and Single Hard Disks

The foremost problem with proper data backup is that it costs money above and beyond the price of your computer, not to mention a fair amount of effort and patience to do it right. The devices that come "free" with your computer that could do the job (i.e. your floppy drive and hard disk) are both fatally flawed for this task. Ever since hard disks grew beyond 10 megabytes, floppy disks became useless for backup purposes (ignoring for the moment that when hard disks were 10mb, floppy disks held less than 400kb). Except for backing up fewer files than you have fingers on your hands, floppy disks are useful only for "SneakerNet" duties; i.e. for moving files from machine A to machine B. Furthermore, floppies get worn out when they are used constantly and have questionnable longevity.

Hard disks, although blessed with greater capacity, hardly qualify as "safe" backup either when you only have one drive. The purpose of backup, after all, is to insure the safety of your data in case of HARD DISK failure. Therefore, aside from "quick and dirty" backups or backups that you can afford to lose (oxymoron anyone?), backing up data onto a single hard disk is almost like having no backup at all. (Backups onto multiple hard disks, however, are another matter and I'll talk about that later.)

With multi-gigabyte hard disks being the norm today, even 100mb solutions - as exemplified by Iomega's (https://www.iomega.com) Zip-100 and Imation's (https://www.imation.com/products/data/content/0,1011,1031,00.html) SuperDisk/LS-120 drives - tend to be inadequate unless you have limited amounts of data to backup. (The Zip-250's are an improvement, but not by much.) In my opinion if you're already prepared to spend a few thousand Baht on a Zip drive, you might consider spending a few more thousand and get something with significantly more capacity and that can be used on other computers in case yours were to fail.

CD Writers

Which brings us to the matter of CD-R/RW's with their 650+mb capacities. While my introduction to CD writers was prompted less by a need for backup than for a need to to dupe CD's, it managed to delay my search for the ideal backup solution for a few years. For most people, I believe CD-R/RW backup is a reasonable solution, especially if one were to get one of the newer models with writing (and maybe even re-writing) speeds around 10x or better. After all, the media is cheap and it's readable by anyone with a non-ancient CD-ROM drive.

I have several reservations about CD writers though:

  • First, sometimes but not always, 650+mb just isn't enough. For me, it might have been fine if I had a single computer, but with each family member having their own computer (and therefore data to backup), 650+mb is borderline sufficient for me. And, with my own hard disk breaking the 10gb barrier, multi-CD backups are now the norm rather than the exception. Being lazy, my preference of course, is for backups that fit onto a single piece of media.
  • Second, as I tend to use CD-R's exclusively (I don't like CD-RW's for some reason), I can't do unattended backups. While CD-RW's can be accessed via drive letters and therefore you can automate backups with them, most CD-R writing programs require hands-on attention. Or maybe I'm just not using the right program. Or maybe I should read the manual!
  • Finally, and I've written about this before, I find CD writers rather fragile, as evidenced by: a) their short one-year warranties and b) the fact that I've waved good-bye to 3 CD-R drives already (2 HP's, 1 Philips). Granted, these were used more heavily than the average home user would and were earlier models, but still this record of failure bodes ill in my mind.

Despite these caveats, I still wouldn't want to be without one, and I make a point of recommending this solution to people who are in the market for backup hardware.

Removable External Drives

Even with a CD writer to meet my basic needs, I've still kept my eye out for some other type of backup which exceeds 1gb. Actually, about 5 years ago when I was still working for my favourite finance company, I heard of a "soon to be available" 4.6gb removable drive from Pinnacle Micro (https://www.pinnaclemicro.com) called "Apex" that would be ideal for backing up our servers. After many delays, the drive eventually dribbled out of Pinnacle several YEARS late, causing untold problems for the company and its management. By that time, though, Iomega had already cleaned up the low-end market with their Zip-100 drives, while offering a higher capacity 1gb "Jaz" drive for those who needed more.

Thwarted by the unavailability of the Apex drive, I ended up getting such a Jaz drive for work, but was aghast by its poor performance. Despite being a SCSI device, the 1gb Jaz was terribly, terribly slow. Add to this the fact that my company was creeping towards it death, and we ended up using it very little. More recently, Iomega released a 2gb version of its Jaz drive. Whether this is faster or not, I'm not sure, but it would be an awful waste if it wasn't.

Actually, there's another removable drive on the market, called the "Orb" and made by Castlewood Systems (https://www.castlewood.com), that I find particularly intriguing. It's similar in capacity to the 2gb Jaz drive (2.2gb for the Orb) but is MUCH less expensive (about US$160 versus US$275 for the internal SCSI models). Not only that, but the data cartridges are cheaper as well ($30 versus $125). The Orb is available in a variety of configurations: EIDE, SCSI, parallel/USB/FireWire ports - some internal, some external, some both.

Speed-wise, the Orb is also pretty impressive. With "verify mode" turned off (granted, not the default setting), it took about 5 minutes to XCOPY 500mb of files from my 7200 RPM Maxtor hard disk to the Orb. Compare this with the 4.5 minutes it took me to XCOPY the same files from one directory to another on the same Maxtor drive. Not bad! And by the way, for Windows and at least for Linux, the EIDE and SCSI versions of the drive require NO special software drivers.

I was so impressed with the specs of the Orb that I picked one up for Christmas and now use it to backup the several computers I have at home. For small business backup needs or when a CD-R isn't quite enough, this is an attractive solution.

Safe Corporate Computing

Since we've touched upon the subject of corporate computing, allow me to discuss some of the data security features employed therein. Companies that use servers (file, or otherwise) tend to employ either RAID 1 or better yet, RAID 5 disk systems. In a RAID 1 setup, each main drive is "mirrored" to a companion backup drive. Any disk "writes" take place identically onto both drives while "reads" occur from either drive. If any one drive were to fail, the companion duplicate would still operate (although it would no longer be mirrored).

RAID 5, meanwhile, employs three or more physical drives. Data written to any drive is also written onto the other drives as smaller "parity" blocks. If any one drive were fail, the data from that drive could be rebuilt from the parity blocks that exist on the other drives after a replacement drive is installed. (Note: The main advantage of RAID 5 over RAID 1 is that it is less wasteful of disk space/drives. Assuming each drive holds 1gb, a RAID-5 setup would require 3 drives to total 2gb, while a RAID-1 setup would require 4 drives. For 4gb, RAID-5 would need 5 drives, while RAID-1 would need 8!!!)

Compared to single hard drive home computers, RAID-equipped servers are much better protected because they have built-in, "live" backups. Most RAID systems also tend to be "hot-swappable" - meaning you can replace a failed drive by pulling it out and sticking in a new one without having to shut down the server. Some of the newer servers even have "hot spares", whereby an unused but nonetheless installed drive is automatically initiated into the RAID scheme in the event of the failure of any drive, mooting the need for manual replacements.

Can you do the above with home computers? Sure you can, but the cost would be prohibitive, especially with RAID 5 configurations which require at least 3 drives, not to mention SCSI hardware which tend to be standard on RAID systems. Of late, though, with ATA/100 IDE disks and controllers approaching the speed of their SCSI counterparts and the availability of affordable IDE RAID controllers such as those from Promise Technology (https://www.promise.com), one could more realistically setup a RAID 1 system on a home computer without going broke buying SCSI hardware. Or a even RAID 0+1 system (i.e. striped and mirrored) if one needed better performance and had more money to burn.

Tape Backup

For corporate IT, though, creating fault-tolerant systems isn't enough. Data must also be stored "off-site" in case accidents or disasters occur "on-site". The most prevalent method of achieving this is by the use of tape backup, which is ideal for several reasons:

  • Many hardware systems - especially large systems - simply don't have alternative means of backup, cost-effective or otherwise.
  • In cases where backup alternatives do exist, tape drives tend to offer higher storage capacities, in smaller form factors which make storage easier.
  • The cost of tapes is relatively cheap especially in situations where large numbers of tapes are required; for example, in "cyclical" backups where dozens of tapes are often used.

Unfortunately, for many PC-based operating systems, tape software must be purchased separately. To make matters worse, tape software tends to be unintuitive, especially given their inability to mount tapes as read/write devices similar to disks. Of course, this is less of a problem in IT environments where skilled or trained personnel perform the tape backup/restore jobs.

Yet another "problem", at least from the individual user's standpoint, is that tape drives tend to be expensive, especially the server-grade models. For businesses, as mentioned, there's almost no choice in the matter but at least they can sometimes use a single tape drive to backup multiple servers. Back in my corporate days, DAT and 8mm tape drives were the norm and yes, these were expensive. More recently, a small company I know was required to get an 8mm Exabyte (https://www.exabyte.com) Eliant 820 drive for their small NT server. The cost in Thailand of this drive was about Baht 70,000+, while the U.S. prices are about $1200. Clearly, one can buy A LOT of computer hardware for that price. And at 14gb, dear readers, this was the "baby" of Exabyte's line of products.

From the standpoint of home users' budgets, these products exist in a completely different world. But this point may be moot anyhow. The advent of Iomega's Zip drives pretty much killed off the market for lower-cost, consumer-grade tape drives, not to mention other removable drive vendors like Syquest and other low-capacity solutions like the 128mb MO (magneto-optical) drives.

This is not to say though, that I never considered getting a tape drive - I did. One drive that was sort of affordable and that could meet my backup needs ad infinitum OnStream's (https://www.onstream.com) "Echo" line of tape drives. While the Echo uses a proprietary type of tape and method of recording (i.e. ADR Technology), I could live with that given that its low-end internal 30gb IDE model sells for only $200 (not dirt cheap, but significantly less than HP or Exabyte drives). Not surprisingly, the Echo comes in a variety of configurations, notably a faster SCSI version, USB, and FireWire, some of which are internal, some external.

But, as mentioned, I ended up getting an Orb instead.

Ad Hoc Backups

Before finishing, I'd like to offer a few more backup strategies. These are decidedly non-traditional cum ad hoc, but they do work (i.e. I've used them before):

  • Peer-to-Peer Backups. If you have a small LAN, you could conceivably "borrow" free space on someone else's hard disk and save your backup there, in a "peer-to-peer" or even a "round-robin" fashion. Safety here, derives from the fact that your data exists in more than one place. Of course, there are a few logistical problems with this approach re: space availability, insuring your backup computer is powered on, etc. But I still do this at home.
  • Backup Server. Although the aforementioned strategy doesn't "scale" very well, it may be "modified" by using a central backup server, which may or may not be "dedicated". This is what we used to do at my old finance company, whereby in the wee hours of the morning all our production servers would send their data to a dedicated backup server. This server was an old 486 tower equipped with five 4gb SCSI drives. Our sole tape drive was attached to this machine, and in the morning tape backup would be performed locally in lieu of backing up "across the wire". (Note: This is hardly a perfect backup strategy though, since data tends to accumulate and the backup server needs to be purged periodically.)
  • Spare Drives. If you have an old or spare drive lying around, it goes without saying that you could use that to back up your data as well. There are several ways to go about this. Myself, I tend to keep my spare drive(s) OUTSIDE of my computer (that is, I don't install it as a functioning second drive) and connect it only when the need arises. This requires opening up my computer's case (my side panel is easily removable and isn't screwed in) and connecting the power and controller cables. A less klunky variation on this is to employ a removable drive caddy, something I haven't tried yet but hope to.

    Not surprisingly, some people choose to install/run their spare/older drive(s) alongside their newer drives. Personally, I don't do this for several reasons:

    • When I buy a replacement drive it's usually because my older drive died or was acting badly or I needed more space. (Here in the U.S. I have another excuse - I see sale prices that are too irresistable - like a 40gb 7200 RPM Maxtor for less than $150!). Continuing to use a flaky drive simply speeds its day of reckoning. I'd rather save wear and tear and semi-retire the drive as a cloned spare. Meanwhile, when I get a new drive, I make sure my replacement is large enough so that I won't need to rely on the older one for day-to-day tasks.
    • With the exception of servers, I don't like installing more than one hard disk in computers. Okay, so I have the Orb for this already (even then the drive cartridge is "out" more than it's "in"). More important though, having two installed HDD's increases heat and power consumption, neither of which are terribly good for desktop computers.
    • Finally, I just don't like having two DIFFERENT hard disks in my computer. Besides the capacity differences, there's also bound to be differences in speed (seek times, ATA-33/66/100, etc.). It just rubs me wrong.

However you look at backup, remember what Nike says: "Just do it!".

Copyright © 2001, Thiravudh Khoman