Jeff: Whether you’re running a pair of drives as RAID 0 or just a single drive, any disk failure is catastrophic. Even if you run a pair of drives in a mirrored config that’s still no guarantee of data integrity if something bad happens.
We’ve had data on whole RAID-5 arrays on our hosting platform rendered utterly useless and unrecoverable because the RAID controller on the last few hours before failure began silently flipping bits, dropping data here, there and everywhere before we detected that it was in real trouble. When it did eventually fail, the data was useless. This happened three times before the manufacturer admitted that there was a problem with a batch of cards.
Fortunately we have a rolling off-site backup that goes back 28 days and we were back up and running at a point just before the wierdness began.
Even if it is just a desktop, if the data is important then good source control, backup procedures and disaster recovery/business continuity practices are imperative no matter how many disks and mirrors you might have. If you have to go sending disks off to data recovery companies then it serves you damn well right for being sloppy with regard to DR/business continuity planning.
I’m curious how much of the performance improvement is due to isolating the boot drive (OS on one drive, apps on another) rather than the speed improvement of the 10000 RPM drive. How much slower is a small 7200 RPM boot drive and a large 7200 Data drive verses the config above?
My understanding is that the 2 independently used spindles provides the perf improvement for VPC use. Is it possible the same is true for non-virtual use?
Is boot time more important than your daily access to your apps and page file? I mean, on an avergage day, how many times do you boot your system vs launching/exiting different apps?
If you have enough memory to support opened apps, you should use your fast drive for apps and not the OS.
Plus you should consider seek and access times and not just RPM speeds.
RAID-0 Actual Application Time benchmarks from May 2006:
“RAID 0 finally shows up to the party and offers a 7% improvement in the Battlefield 2 scores but otherwise does not offer any tangible benefits, and it even posts slower load times in the Oblivion and Half Life 2 benchmarks.”
Now ask yourself: is that worth doubling the risk of losing all your data?
Benches on my machine showed an almost perfect doubling of drive throughput, and timed load times of zones in WOW (using a plugin to time zone loading times) resulted in something like 25%-50% faster loads on zones, and a similar speedup in boot times on XP.
The trouble with the benches you cite is that in things like loading a zone in Quake, only a fraction of the level load time is spent loading the level off the disk, so you end up hitting Amdahl’s law. They also compare a 10k drive vs a RAID array of two slower drives.
You also didn’t notice what I said, which is that I have no real risk of data failure, since I have automatic backups to a third drive on my system. If I was building this from scratch, it’d be RAID5 all the way, but the marketing on the box of the mobo tricked me, and so I have an ad-hoc RAID5 with a RAID0 array backed up by a single drive.
Also, I highly not recommend buying an off the shelf Dell or whatever at your computer store. The reason they’re so cheap is they cut corners where you’re not looking. I can’t count the number of times I’ve popped open a premade computer and seen a single RAM slot, or no expansion PCI slots, or only a single IDE connector, or whatever. When my fiancee wanted a new computer, I went crazy trying to find a computer that had everything we wanted (using all the build-your-own websites, like Dell’s), and they either couldn’t do it, didn’t have it, or were grossly overpriced. So I went to Fry’s and spent the afternoon building a computer for her. Not top of the line, but at that price-inflection point where things start getting cheap, RAID1 for safety, all came in under a thousand bucks. Runs like a champ.
My computer has a single 80 gig SATA 2 hard drive. No problems to date. It runs a Pentium 1 processor that runs nice and cool. I also make extensive use of memory sticks for temporary storage. They’re an excellent substitute for floppies.
I simply like to keep things simple.
I’ve been looking at upgrading the hard drive in my desktop for a while now. It got pushed to the back burner when I got my laptop, but now I’m looking at this with interest.
Jeff, you mentioned that you use sorbothane and eggcrate foam when mounting your drives…
I was wondering if you could go into a little more detail about how you use these materials to dampen HDD noise and vibration in your rig. Perhaps a new “How-To” article?
I ask b/c I have a Dell Optiplex GX620 w/a pair Seagate (7200rpm) SATAII drives and they vibrate so much that at times it seems the freaking case will fall apart. So I’m looking for some tips to calm these suckers down and bring the box (any my office) back to an acceptable decibel level.
It’s funny how most of the commments are just “blowing the same horn”. While the one comment that wasn’t agreeing with Jeff got completely ignored.
Well here’s my 2 cents. This whole article basically advertises the 10k rpm Raptors as the ultimate solution. Which they’re not.
Synthetic benchmarks aside, what do you really profit? Maybe your OS boots a few seconds faster and your programs run 10-15% faster? So what? In a matter of days you grow accustomed to the “speed improvement” and you’ll (again) want more. I’d say that’s normal human response…
First of all, for about the price of one 74GB Raptor you can get a 400GB Barracuda. I can think of better ways of spending my hard-earned cash. I’d say RAID will give you more in terms of data security as well as some speed increment. And you don’t have to spend a fortune to have it… as for myself, I still rely on a single 120GB Barracuda to do everything I need. I backup my data to an external hard drive every two weeks. The odds of two hard drives dying at the same time is very very small. And if there’s a fire or a quake, your hardware is toast anyway no matter how many hard drives you got in there
Yeah, it’s $500 each, but it’s the BEST $500 you can spend on your computer… Additionally, it’ll be quite a bit cheaper if you just wait 6 months (its still fairly new).
As president of Puget Custom Computers, I get a unique perspective on computer products and technology. Our company specializes in selling high performance custom computers, and that naturally brings up the question of RAID often. There is an overwhelming opinion out there that if you have the money and want a blazing fast and stable computer, that you should put your hard drives in RAID. We have known for years that this perception is just flat out wrong, but the problem is that the idea is so widely accepted that it is nearly impossible to convince our customers otherwise. In fact, if we try too hard to talk them out of it, we end up losing the sale! So, should we be selling configurations that we know are flawed, for the sake of making the sale? To be honest, if it comes down to it we will, but not without a fight! This article is just the latest effort in educating the public about RAID.
Heh… Considering my PC is not my main machine, and all it has on it is games where the user data is stored remotely (like warcraft), and because I’m cheap, I’m using:
A Mylex DAC960 triple channel SCSI hardware RAID card I got for $5.
And six 9gb 10k rpm Seagates. (all discarded parts from their old rigs) Two drives per channel, RAID-0. 54gb for $5.
I don’t care about the reliability cuz the data on it isn’t important, I just care that the system’s fast. And it’s just plain unbelievable.
Yes I too am running a WD Raptor 10,000 rpm drive. I installed Linux FC6 the day I got the drive and I’ve felt like I have a new computer ever since! I highly recomend a 10,000 rpm drive to anyone looking to make theyr’e comeputer dangerously fast.
Very true, I put together my new system with a 10k raptor and will never have another drive less than 10k again, until SSD is cheap.
I ghosted my old drive (the fastest 7200 at the time, a Seagate 7200.10) to the 10k raptor and couldn’t believe the difference. This is the exact same install and computer, my boot times went from around 30+ seconds to under 15 seconds.
I used a HD silencer to quiet it down, its now even quieter than my old seagate and runs at about 33C, very nice.
Also stay away from the windowed versoin raptor, the raptor x, it runs hotter, and is much louder than the normal raptor.
if you are truly looking for awsome speed thats a bit out of the realm of an actual hard drive… look into I-RAM its a slick idea that im amazed hasn’t really been done much with…
The Raptor drives certainly are the fastest SATA drives you can buy, but the performance margin over a fast 7200 RPM such as the Hitachi 7K500 is pretty minimal. In the performance tests you linked to (e.g. StorageReview) the Raptor had an advantage of 12-21% in single user benchmarks. Since the HDD is just one part of overall system performance, that likely translates into substantially less than 10% overall performance increase as seen by the end user, which is small enough that in most circumstances it is at or below the threshold that a human can reliably detect.
What that means is that other than running a benchmark or doing something that is very IO intensive (and in which the HDD is the performance bottleneck, not RAM, CPU, memory aperture, etc.) a typical person sitting in front of two otherwise identical systems, one with a Hitachi 7K500 and the other with a Raptor X, most people could not tell you which was which. As an example the performance increase for a CPU before people can perceive a difference is over 20% (with apologies to all those people who get so excited when their overclocked systems score 12% higher on benchmarks).
Bottom line - it can’t hurt to use the Raptor, as it’s an excellent drive and appears to be very reliable. Just don’t get caught up in the comments above. It’s unlikely that most people, including gamer and power users, will notice much (if any) difference moving from a fast 7200 to a 10K drive, and the likelyhood that the difference will be “dramatic” is vanishingly small. So why do people post comments like that? Look up “confirmation bias” on Wikipedia - people want to believe that their choices have had a very positive effect (the mechanism is not unlike the placebo effect).
is that other than running a benchmark or doing something that is very IO intensive (and in which the HDD is the performance bottleneck, not RAM, CPU, memory aperture, etc.)
Like, say, double-clicking on an application icon to load it from disk into memory? Something most people do dozens of times each day? Disk IO is a significant bottleneck in today’s PCs. Yes, you should have as much memory as possible, but you will be inevitably accessing the disk at some point-- and all disk activity is 10-30% faster on a 10,000 RPM drive.
As I said in the entry, I was a skeptic too. I first noticed the difference while using someone else’s machine.
Barry, if you don’t believe me, buy the drive from some place with a return policy and try it yourself. I don’t think you’re basing those comments on actual experience.
