here was a supermarket in my hometown that used to drive me crazy. They'd advertise low, low prices--real bargains, compared to competitors. But then, you'd notice this small note at the bottom of the ad: Plus 10%. (Or worse, you'd discover it when you went to check out that "We have to add 10% to your total.") So what started out looking like a great bargain turned out to be not much different than you would have gotten at any of the other stores.
Now, I wouldn't have minded so much if that 10% actually meant something. Like, "We just show our wholesale price on the shelf or ad," and then they have to add overhead and profit on top. That would explain the 10%. But that lower price wasn't their wholesale price at all. It was just a number set by the store. So what was it there for, if not to try to mislead you while you shopped into thinking that that would be the price you would pay?
We all hate finding the "gotchas" in modern advertising, the little asterisks that indicate your mileage will vary. Sometimes it can't be helped. But many times it can be helped, like when someone offers you a free week in Hawaii but you have to buy a high-priced airline ticket to get it. Don't be surprised when the ticket is the same price as the air/hotel package would be from a reputable agent.
Unfortunately in our IT world, we suffer from the same thing. But rarely do we even get the subtly placed asterisk that at least sets the suspicious shopper on guard. Too often we are left with unqualified, misleading performance numbers that don't help us in making an educated choice in the market.
For example, a recent Microsoft press release loudly proclaimed that "Windows 2000 Professional is the fastest Windows yet." It supported this with independent tests from Ziff-Davis Labs. Certainly seems straightforward enough. Seems like a solid vote for the newest MS OS.
However, the Ziff-Davis report doesn't say this. In fact, the report actually draws the opposite conclusion. It states, "We found that Windows NT 4.0 provided slightly better performance than Windows 2000." Now, we could quibble about the added advantages of Win2K as a platform, but there can't be much argument that the ZD report says one thing while MS proclaims another, citing the same. Who does that kind of hype help?
In our optical world, we suffer consistently from the deceptive nomenclature attached to CAV CD-ROM drives. No consumer I know understands what the "Max" in "40X Max" means. Why can't we simply state the drive's average speed? If the drive only reads at 40X the last available 3MB of a CD-ROM disc--where there's rarely any data--why can't we explain it better as 15X with a Burst Mode to 40X for the last 3MB?
Every gaming fan to whom I have shown the Kenwood TrueX drive adores it [see review, January 1999, pp. 58-61]. Gamers love the consistently high throughput that the Kenwood provides. Not one wanted to go back to the cheapo drives with their 15X mean performance. But until I showed them the drive, only one understood what TrueX was or why those Brand X CAV CD-ROM drives don't give a full 40X.
Finally, where are we with DVD-RAM numbers? One would think from reading vendor info that DVD-RAM drives write at a full 1.35MB/sec. But that ain't so.
While it may be possible to write a premastered image file to DVD-RAM at that speed, most files aren't written at that speed, and most file systems don't typically use 2.7GB files. Most files are much smaller; in that case, the drive is much slower.
It is slower because it takes longer to do many small writes with DVD-RAM (or CD-RW, for that matter). In Windows, Mac, and Unix, writing to disc involves not one but two operations: recording the data, and recording the data's location in a table at the start of the disc.
The optical write head is much heavier than the magnetic one in a hard disk, so seek times are much slower (by almost a factor of 10). Since each recording involves two seeks--one to record data and one to record the data's location--this slows the recording process considerably. Since data is recorded in small blocks, there are many seeks for every file. This means that recording DVD-RAM must be farther off its advertised speed than magnetic. But by how much?
In our tests, using a Toshiba SD-W1101 DVD-RAM drive on a Windows workstation, we weren't able to do better than 450-500KB/sec, barely 1/3 the write speed spec'd for the drive. (It didn't matter much in the results what block size we used, although at 2K the RAM drive bottomed out at 250-300KB/sec.) This is significantly less than the 1.35MB/sec figure that one might be counting on when buying a drive.
It isn't that we'd like to see lists of asterisks on the side of every box. Nor do we expect vendors to display the worst performance possible for their devices either. But it would be nice when we invest in a given technology to know what the performance ratings mean. After all, if Linux, Windows, and Mac file systems all record similarly, wouldn't it be enough to look for solutions that record at a given figure, at least for those platforms?
While standalone users might tolerate poorer results from the latest technology, few network administrators can. Not when the purchases are for hundreds of units and the complaints might be in the hundreds as well. We simply won't get a budget next year if that's the case.
The Network ObServer columnist David Doering (firstname.lastname@example.org), an EMedia contributing editor, is also senior analyst with TechVoice Inc., an Orem, Utah-based consultancy.
Comments? Email us at email@example.com.