LinuxWorld posted an article today on how to install another hard disk in Linux. The guide’s pretty good from the software side.
The advice is slightly questionable from the hardware side. Author Joe Barr states that it doesn’t matter which connector on the cable you use, as long as one drive is jumpered master and one drive is slave. For years that was true, but you’re actually supposed to put the master on the end and the slave in the middle. Usually it doesn’t matter. But the newer your drive is, and the newer your controller is, and the longer your cable is, the more likely it is to matter. You also shouldn’t attach a drive to the middle and leave the top connector hanging. Again, you can usually get away with it–and people have gotten away with it for more than a decade–but the likelihood of not getting away with it increases with every passing day, as hard drives get faster and faster, and thus more and more touchy.
What happens when you do it wrong? Usually it works anyway. Sometimes it’ll be flaky. And sometimes it won’t work at all. Don’t you love predictability? So it’s really best to follow the rules unless the layout of your case makes that impossible.
But the main reason I’m writing is because the usual expected flamewar erupted in the discussion thread. Barr bought a Western Digital drive. Predictably, someone responded that Western Digitals are junk. Then someone responded to the response and said Maxtors are junk but Seagates are good. Then someone responded to the response to the response and said Seagates used to be junk. Before you knew it, every brand of hard drive on the market–IBM, Samsung, Seagate, Maxtor, Western Digital–had been trashed. Curiously, except for Fujitsu. But Fujitsu recently had a big scandal with a failure rate on one particular model of drive higher than 90 percent. (Meanwhile, my own experience tells me Fujitsu SCSI drives are fantastic.) And lately, Samsung drives have been getting praise all over the place. So what gives?
The problem with these statements is there’s a degree of truth to all of them. There was a time when Maxtor hard drives were the worst thing you could buy. Ever heard this joke? Fast, reliable, and cheap: Pick two. Well, in the early to mid-’90s, Maxtors weren’t fast, they weren’t reliable, and they weren’t consistently any cheaper than any other brand. The only reason to buy them was because the familiar red boxes were everywhere. The only place you couldn’t buy them was the corner gas station. Well, in St. Louis at least.
During the same time frame, Seagate had similar troubles. Their drives were expensive, but they weren’t fast. I didn’t see enough of them to get any kind of handle on reliability because I was so turned off by their price and underachieving performance that I wouldn’t go near them, and neither would anyone else I knew.
In the mid to late ’90s, it was Western Digital’s turn to go 0 for 3 on fast, reliable, and cheap. From 1997 to about 2000, I saw more dead Western Digitals than every other brand, combined. And I saw a lot of drives come across my desk.
With its GXP series a couple of years ago, IBM had the fastest drives on the market, and they were also among the cheapest. But they were exceedingly touchy, and became notorious for premature failure.
I bought a handful of Samsung drives over the years, never willingly, because of their terrible reputation. They’ve been reliable. And when you look at reviews of their recent drives, they run cool and they’re reasonably fast. They’re not necessarily the fastest on the market at any given time, but they may very well be the best combination of fast, reliable, and cheap right now.
I’ve been around long enough and seen enough that every time I see unqualified statements like “Western Digital drives are junk,” or “Maxtor drives are junk,” or “Seagate and Maxtor drives are the best,” whether it’s from some end user in a discussion forum or a professional hardware reviewer, I get suspicious. The end user is probably basing those conclusions on a too-small sample size, and the professional reviewer probably isn’t doing the necessary homework.
Let me tell you why.
We know how to build a completely reliable hard drive, one that will run for 10 years and never have problems. But it would cost too much money, its capacity would be too small, and it would be too slow. The technology in hard drives changes with each generation, and the company with the best technology is generally the one that produces the most reliable drives. But the most advanced technology isn’t always the best technology, as IBM found out with its GXPs. The GXPs were too far ahead of their time.
It should come as no surprise that when Maxtor was producing junk drives, they weren’t in very good shape financially. There wasn’t much money for R&D. When Maxtor’s financial situation improved, its R&D improved, and its drives became faster and more reliable.
There was a time when someone could ask me what hard drive to buy and I could give them a brand and model number that would give them the best combination of fast, reliable and cheap. But my newest computer at home was built in the summer of 2001 and I very rarely work on desktop systems anymore–I’m a server guy these days, and I have been for the past 18 months. If I’m honest with myself and with the person asking the question, a lot can change in 18 months. In 2001, as far as I could tell, the best drive to buy was a Maxtor and the worst to buy was a Western Digital.
I can go with my old prejudices and continue to dispense that advice indefinitely. But there was a time when that was reversed. And what about Samsung? They’re quiet and they run cool, which is a good sign, they’re very affordable, and while they’re almost never the fastest, they never get blown out of the water by benchmarks.
The best thing to do is to talk with someone who actually works with the equipment on a regular basis, and in large volumes. I want the opinions of someone who speaks from recent knowledge and experience, not someone speaking from old prejudices or a gravy train of free hardware. That means I’d call up a couple of former coworkers who still do some desktop support, or who at least handle the RMAs for subordinates who do desktop support. I’d ask them whose drives have been failing the most lately, and if they notice much performance difference between brands. Benchmarks are more precise, but they can also be fooled. If you can’t notice the difference in the real world I really don’t care about it. If you do notice the difference, I don’t care much about percentages. It’s subjective, but as long as I trust the people whose opinions I’m soliciting, that doesn’t matter much to me.
And after talking to a couple of people who actually handle a few drives a week, I’d go plunk down my cash.