Advances in New Western Digital Drives

An anonymous reader writes "The Western Digital Caviar SE16 WD2500KS 250 GB hard drive has 300 MB/sec transfer rate the drive has a monster 16 MB cache, both of which should make it one of the best performing 7200 RPM drives on the market. WD categorizes this drive in the "Highest Performance" section of its desktop market, so its safe to assume that is has solid performance without the expense of an enterprise level drive. With products like this available, advances are being made in the storage industry that are not being rivalled by those in other areas of computing, especially considering the price level of this drive."

Comments

This is not new or special

[Anonymous Coward]

WD released this drive at least 3 months ago, and other drives with 16mb caches have been out even longer.

This is just another useless anonymously submitted article by Sal Cangeloso that may in fact be a slashvertisement. Notice the price listing on the first page, unless of course you have your ads blocked.

Re:This is not new or special

[MoonBuggy]

Not only is it an advert, it's an incorrect one. If you want to be using an "enterprise level" drive, it's the 400GB WD4000KD you should get - same series (Caviar SE16), but the hardware is physically identical to the newest 400GB 'Raid Edition 2' from WD. The 400GB Caviar SE16 model is based on the 10k Raptor family of drives designed for maximum speed, whereas the 250GB SE16 is descended from the standard-issue Caviar family. The only place where the 250GB model beats the 400GB is the support of a 300MB/s SATA2 bus rather than the 150MB/s of SATA1, but since no drives can actually deliver anything like 150MB/s transfer, it's redundant anyway.

Note that I don't intend to advocate any one of these drives - I couldn't care less what you buy, I just want to lay out the facts properly.

Re:This is not new or special

[ergo98]

Wow.

Clearly it's a slashvertisement, as all of the linked articles are

• Terribly, terribly, terribly written
  
• Incredibly abusive of readers, spreading the limited, technically vacuous horrific prose over a dozen pages
  

If the Slashdot crew accepted those submissions without payment then they should commit hari kari now, because their use on this planet is done. If they did receive a kick-back - which I think is unquestionable - then I think this pretty effectively puts them on notice.

Good catch.

Remarkable that Slashdot is at such a vulnerable time, when there is a tide of credible competitors emerging, and they pull stunts like this.

Re:This is not new or special

[spatenbrau]

According to the article, the 250GB bursts hit 171 MB/s, so actually it would be hindered by SATA1.

Seeing how the overall data-rate off of the heads is only in the 60MByte/sec to 90MByte/sec range, all this talk of 300MByte channels is bordering on dishonesty with numbers. The burst rate sounds like it is simply the speed at which the on-board cache can be read at. That isn't going to a number that influences much other than artificial benchmarks.

This article is just another article from an ever growing number of "fan-boy" review sites that read like ad-copy. A real review would test the disk throughput with something similar to:

dd if=/dev/wd0c of=/dev/null bs=1m

and then note the true MBytes/sec that the disk achieved. This is the best-case number that essentially allows the disk to stream with very little track-to-track stepping. Good disks will be able to do 60MBytes/sec on the outer tracks, but will usually slow to 2/3rd's that speed on inner tracks.

Re:This is not new or special

[rew]

could two (or more) SATA1 drives on a SATA2 bus exceed 150 MB/s in total? I would think not, in which case SATA2 is a big advantage if you want multiple drives on a bus.

Well...... I'd vote against two drives on one bus. SATA is point to point. Max one drive per connection.

So the article, and the WDC website claim 300MB per second transfer rate. That's THEORETICAL buffer-to-host. Apparently someone measured that at 170 Mbyte per second.

Platter to buffer is an impressive 748 Mbit per second. That's an impressive 93.5 Mbyte per second!!!!! Read on and "sustained max" is 65 Mbyte per second. Duh. That's pretty normal for a 2005 drive. Nothing out of the ordinary.

Now about the advantage of SATA2 (300MB/s) versus SATA1 (150MB/s).

You get to use the extra bandwidth if you let the disk buffer (16 MB) fill with data from the platter (65Mb/sec) -> about 250 ms. And then suddenly you decide that you need all that data in the computer's main memory.

If this happens just once, you win some 200ms. That's hardly noticable. It would matter if it happens lots of times in a row. But it is a very very funny traffic pattern where your disk is able to cache 250 ms worth of data into its buffer while the computer doesn't "see it coming", and has to take full advantage of the much smarter drive. Right! Nah.

The buffer should be large enough that if you're streaming say 200Mbyte per second off 4 drives in striping RAID, the drive should "stay ahead", and while you're handling the other three drives, the buffer shouldn't overflow. So the RAID block size should not exceed about 10 mbyte. Well, that's outrageous. The buffer is QUITE large enough.

At 7200 RPM, or 120 rotations per second, and 65 Mbyte per second. we can calculate the data density to be about 0.5 Mbyte per track. So with a buffer of more than two tracks (one megabyte), you quickly approach the theoretical throughput.

Really, there is nothing special about this drive.

Re:This is not new or special

[Shanep]

According to the article, the 250GB bursts hit 171 MB/s, so actually it would be hindered by SATA1. Burst speed isn't my #1 consideration anyways, but it's something.

Drive quoted burst speed comes from or to on drive cache anyway. That cache is good to help the drive sustain it's highest read transfer rates through read-ahead (when the OS comes back for the next block, it has already been read from disk) and also an OS can send small writes to the drive faster. But in practice this mostly just helps a disk to meet it's highest sustained transfer rates. The burst speed sounds good and 16MB cache sounds good, but in these modern times, when we use OS' which use free memory as buffer/caches, we have a LOT of memory and that memory is REALLY FAST, on-drive caches are mostly being used as buffers. As far a caching goes, they don't really get used all that much, since re-reading a block will almost always come from system RAM before it comes from drive cache RAM. Sure it is true that the read-ahead caching on the drive is caching, but in practice it is mostly used as a buffer.

Are there some other SCSI drives with higher performance now?

From a practical point of view or from a meaningless burst speed point of view due to large on-drive caches and fast busses?

I have a Fujitsu SCSI320 drive which sustains about 94MB/s at the beginning of the disk, which slowly tapers off to about 64MB/s at the end of the disk. That is faster than the raptor and this SCSI drive is also faster than the raptor in other aspects like read service times, I/O rates, etc.

There have been Fujitsu, Maxtor, Seagate and Hitachi SCSI drives faster than the raptor for a long time. The Maxtor Atlas 15K II is really fast.

In fact, as far as sustained reads and writes go, access times and sustained I/O, has SCSI EVER lost the top spot?

Re:This is not new or special

[Not The Real Me]

SCSI has always held the top spot.

This IDE has a buffer to disk speed of 93.5 megs/sec. Interestingly enough, the article fails to mention what the CPU load is.

For all important items, I use a SCSI, usually Ultra160 since U320 is still quite expensive. IDE is what I use for secondary storage since IDEs do chew up CPU cycles.

Re:This is not new or special

[timeOday]

The burst speed sounds good and 16MB cache sounds good, but in these modern times, when we use OS' which use free memory as buffer/caches, we have a LOT of memory and that memory is REALLY FAST, on-drive caches are mostly being used as buffers.

I've questioned the usefulness of hdd cache compared to OS main memory cache before on slashdot and gotten flamed. Unfortunately I've still never seen any benchmark that convinced me of whether large onboard cache really helps, or just helps results on benchmarks which intentionally avoid OS disk caching. If anybody has some hard info, post a link.

In fact, as far as sustained reads and writes go, access times and sustained I/O, has SCSI EVER lost the top spot?

Sure, SCSI drives always had the quickest seek (access) times. The highest RPM drives were offered only with SCSI interfaces. That's what set them apart as "server" drives. I don't think there have been any 15K SATA or PATA drives, have there?

Re:This is not new or special

[Shanep]

I've questioned the usefulness of hdd cache compared to OS main memory cache before on slashdot and gotten flamed. Unfortunately I've still never seen any benchmark that convinced me of whether large onboard cache really helps, or just helps results on benchmarks which intentionally avoid OS disk caching. If anybody has some hard info, post a link.

I don't have a link at the moment with any hard info. But I did recently test re-reading a 1GB file in FreeBSD 6.0 Release on my AMD XP2800+ with 2GB DDR ram...

came out to about 588 MB/s.

(that's 2^20 MB/s and not the bogus 10^6 MB/s.)

I might be able to dig up a link which proves that larger on-drive caches can actually hurt performance, since I did see a benchmark at Storagereview which showed this in two almost exactly the same WD drives which only differed in cache size (same model drives, except one was a "Special Edition" with larger cache only)....

Nice ad

[Anonymous Coward]

I wonder how do paying subscribers feel about seeing ads before everyone else!

Re:Nice ad

[Phroggy]

We're not especially thrilled about it.

YAPR

[legLess]

Yet Another Press Release. Nice to see that Taco's tight editorial control hasn't been impaired by too much turkey. The guys at XYZ Computing are giving each other high-fives right now.

Thank you

[agrippa_cash]

Thank you Slashdot, for bringing to my attention this exciting new service or product!

Advances that aren't being rivaled?

[Anonymous Coward]

Is the poster serious? Hard drive performance is one of the slowest areas of advancement in PCs there is. Granted that there's legitimate reasons for that, but to say that because its got a bigger cache we're seeing advances not seen anywhere else is laughable.

Compare a video card from today to one two years ago, and do the same thing with hard drives. The amount of "advancement" in the video cards far outpaces the drives, except for the really big drives that can store weeks worth of pr0n at once.

Re:Advances that aren't being rivaled?

[Rosco P. Coltrane]

Compare a video card from today to one two years ago, and do the same thing with hard drives.

I'm trying, but really the VGA plug won't fit the IDE connector. I'm so confused...

Re:Advances that aren't being rivaled?

[Grishnakh]

I don't see it this way. When I look back over all my computers from 1991 to now, hard drive performance and capacity seems to be the most obvious thing that has increased. Back then, 40MB drives were common, and it took a long time to read all that data. Now, I can read that much data off my hard drives in less than a second, and 400GB drives are becoming common--a 10,000x increase in capacity. Back in the old days, I really had to worry a lot about what I had stored on my HD, because the space was quite limited, even with the smaller software of the day. Now, I don't even have to think much about it unless I'm storing lots of movies.

What other factors in computers have increased so much in that time?

Network speeds? Nope. Back then, I had a 2400bps modem, and 9600bps was fairly common. Now, the most you can have is 56000bps (theoretical), only a 23x increase. Pathetic. Good thing we have broadband now, but that's an entirely different thing. For LAN networks, we had 10Mbps ethernet back then; I'm not sure if 100Mbps was out by that time or not. Now, 100Mbps is common for most places (home networks, etc.), and GbE is getting common, though it's probably used a lot more in datacenters. That's a 100x increase.

CPU? Well, we now have 3.8GHz P4 processors (which perform about as well as much slower Athlons...), and while they're certainly much faster than my old 20MHz 286 in 1991, I really don't see much difference in regular desktop usage between my 3.6GHz P4 and my old 1GHz Duron from 3-4 years ago, or my 300MHz Celeron before that, unless I'm running transcode or something (i.e., not often). Word processing, web browsing, etc. all looks about the same to me, although I can open more browser windows now with 1GB of RAM which was uncommon 5 years ago. Furthermore, CPU performance has now hit a brick wall, and multi-core is the only way past it, which isn't helpful for single-threaded apps.

Local non-HD storage? Back in 1991, we had 1.44MB floppies, and now 4.7GB DVD+/-R discs are common, a 3,264x increase. I'd call this probably the second most noticeable and usable improvement in computers besides HDs, and interestingly enough, these are also mostly mechanical in nature, not silicon-based.

Video cards? Back in 1991 or so, I had a video card that could display 1024x768. These days, the highest resolution most people run is 1280x1024, not much of an improvement, and the highest is usually 1600x1200. Whoopee. Monitors have gotten bigger for most people, but they had huge monitors back in 1990 too, except they were really expensive. Of course, now video cards have entire processors on them specialized for 3D graphics, but what good is this? If you're a normal computer user, it's pretty much useless (no, teenagers who spend all their time playing video games are not "normal computer users"). GPU technology is great for console systems, but this discussion is about PCs, not video games.

Therefore, I'd say that hard drive performance is definitely the most obvious, most important, and most usable area of advancement for PCs over the last 15 years.

Re:Advances that aren't being rivaled?

[vrt3]

CPU? Well, we now have 3.8GHz P4 processors (which perform about as well as much slower Athlons...), and while they're certainly much faster than my old 20MHz 286 in 1991, I really don't see much difference in regular desktop usage between my 3.6GHz P4 and my old 1GHz Duron from 3-4 years ago, or my 300MHz Celeron before that, unless I'm running transcode or something (i.e., not often).

You don't see much improvement in regular desktop usage precisely because all the other components haven't had performance increases as much as the CPU. Memory is notoriously falling more and more behind. And many tasks are I/O-bound rather than CPU-bound, which means that speed increases in the CPU won't help unless the I/O bottlenecks get faster. One of those bottlenecks is in many cases still the HD.

Interesting Fact

[matr0x_x]

A little known fact about the WD2500KS is that it has a sister WD2800KS due out in 4 months with double the storage and 35% higher performance. Of course the cost isgoing to be MUCH higher too

Re:Interesting Fact

[discord5]

Not if they can plug it on slashdot for free. 30% marketing reduction cost guaranteed.

Anonymous Coward posts Slashvertisement

[Yahweh Doesn't Exist]

nothing to see here.

desktop hard drives are quite possibly the most boring technology possible, except maybe non-wireless network cards. who cares?

Oh puhleeese

[$RANDOMLUSER]

A 3-platter 8.9mS seek-time 7200 RPM drive with a 16MB cache? You better use this puppy for video files, 'cause the only thing that's more tuned for sequential access is a tape drive.

File this under "Ads that matter".

Big, Slow Drives

[jimmyhat3939]

Note to hard-drive manufacturers:

Please come out with a larger, slower drive for those masses of us who want to store very large quantities of data but don't care so much about 7,200 RPM or large cache sizes and whatnot.

When will the 1TB hard drive come out? When oh when?
--
Free 411! 1-800-411-SAVE

Meta-Comments

[TubeSteak]

1. I don't like the warranty
2. I've had bad experiences with WD drives
3. I've had great experiences with WD drives
4. 250 GB isn't really 250 GB*
5. This review isn't comparing similar drives
6. My RAID array is faster
7. RAID-0 isn't really redundant

And my quick summary of the aritcle:
$125 (50 cents per GB)
SATA
Not the fastest drive on the market

*In this case, the formatted drive really does hold 250 GB

Slashvertisement.

[imsabbel]

Its getting more and more annoying...
So this drive is great... says WD.
So obviously is MUST be great.

And i really like reading that it has a 16" monster cock... ^h^h^h^h^h^..^h 16Mbyte monster cache. You can really feel the journalistic integrity OOZING out between the letters. I mean, thats SOO great considering that currently my windows uses 360Mbyte as file cache, connected with 6.4Gbyte/s.

And a 250Gbyte drive is SOOOO revolutionary. I mean, thats the smell of the future. Almost as if we were already in the 3rd millenium.... oh wait, we ARE there, and drives of this size have been around for 2.5years+ already.

And Sata-2 transfer limits are SOOOO useful as a dazzling number when your drive barely reaches 70Mbyte on the outermost tracks for the first Gbyte.

Product quality

[Mr Smidge]

Great! Now when can we expect similarly advanced levels of production and refinement in the spelling and grammatical skills of our summary writers?

300MB/s my arse

[Darren Winsper]

That drive uses SATA 300MB/s, which means a peak speed, not a sustained speed. It seems the drive can manage 50-60MB/s sustained.

WD hard drive quality: you get what you pay for

[digidave]

Western Digital SE drives are consumer-level drives not known for having high quality.

WD also sells IDE and SATA RE and RE2 enterprise drives with MTBFs of 1 - 1.2 million hours. Why would anybody want to halve the MTBF of their drive by getting an SE drive just to save $30?

Their RE and RE2 drives (or Raptor if you don't need huge capacity) are very high quality. These drives really kick ass and come in 8 MB (RE) and 16 MB (RE2) cache models. I bought four of the REs for a server and they've been performing flawlessly.

Where have you been living, in a bubble?

[Anonymous Coward]

"With products like this available,"

Yup, drives like this have been around for the last 6-12 months. They've probably shipped tens of thousands of them and you think they're cutting edge?

"advances are being made in the storage industry that are not being rivalled by those in other areas of computing,"

Not really, have a look at the access time - 8.9 ms - this drive is just as fast as one from 8 years ago, it's just bigger. And guess what? that's why it has a 16MB cache. More platters, more heads, more cache plus greater data density equals... same access times. Hard drives don't scale up as well as other technologies.

"especially considering the price level of this drive"

Hang on a second, you can get cheaper than this. You can also get WD Raptors, which although smaller in capacity, are much, much faster. In fact, this is just a hard drive, like many other hard drives.

These are the stories I hate. Pointless, heartless drivel passed by the editors who well, don't really edit, and appear to be out of touch with their readers, not to mention their market segment. An absolute, total and utter waste of screen inches - the kind of crap I'd expect to spout forth from a zit-faced store assistant who didn't know a molex connector from his arse. An embarrassment to read on Slashdot really. Shame on you.

From TFA

[SteWhite]

And if you spend 30 seconds looking at the article, as CmdrTaco should have, you will see that this drive does not deliver 300 Mb/sec. As reported by SiSoft Sandra, it gives 52 Mb/sec. Which many other high performance drives can match. The 300 Mb/sec figure is cache to host transfer speed, which with a 300 Mb/sec transfer and 16 Mb of cache, could be sustained for a whopping 0.0533333... seconds. Wow.

16 mb cache eh?

[TheSkepticalOptimist]

Had it on my Maxtor for the last 3 months. Way to catch up with the times.

But in general, most hard drives are still severely underperforming, regardless of their specs on paper. Its the single biggest bottleneck on today's systems, causing system hangs and stutters on even the fastest systems.

This industry needs a kick in the ass!

300mb/s transfer rate on a system capable of procssing 8GB of data per second, that is nothing to rave about. Also, most systems still work off the principle that you can only read or write one operation at a time. Sure caches offer speed improvements, but there is no reason why a hard drive can't have multiple read/write heads to access different sectors of the disk as the same time, turn a SINGLE drive into a STRIPE set. It can be done if you put your mind to it.

In order to get decent performance out of hard drives today, you have to buy 3 or more drives and setup a RAID 5 system and waste one drive for redundancy. But of course, the hard drive makers want you to do this, buying 3 drives instead of 1 to get decent performance is their goal, I am sure.

The problem is, hard drives have reached a point where they are cheap commodities. The leaders in this field have long since figured out how to make cheap and reliable drives while increasing storage space proprotionally to a reduction of price. Prices for hard drives keep getting cheap, and hard drive makers are not making that much money off them anymore.

This is why the whole external hard drive fad has started up, taking a $50/unit hard drive, stuffing it into a $25/unit external enclousure, and charging $300 retail for it. Add a back up button and maybe a USB port on the front and you suddenly create a "Must Have" product.

Because of this, there is no motivation to improve the technology. They figured out a few years ago how to dramatically increase storage space breaking what they preceived to be a physical barrier, and since then hard drive storage capacities double every year. With the ability to stack bits on top of each other, we will hit terrabyte storage capacities next year easily. But they are using the SAME TECHNOLOGY, the same magnetic media, read/write heads, packaging, I/O boards and chips, cache, etc, etc, etc. They may tweak these componenets to work with higher capacities, or improve performance somewhat, but nothing has really changes in terms of how hard drives are manufactured and designed. Compare this with the CPU industry that re-invents itself every 18-24 months.

Where are solid state hard drives? Where is my obscenely fast gamer performance extreme drive capable of feeding data to my video card in real time? Where is my tiny thumb sized drive capable of storing terrabytes of data?

The hard drive industry is just dolling out minor improvements and tweaks to existing technology, and expecting to be slapped on the back every time they boost storage capacity by a third or transfer rates by a quarter. Something seriously needs to happen in this industry to make hard drives hot technology again, its grown quite stale and cold over the last 10 years. Until then, I am underwhelmed with these kinds of press releases. Way to go Western Digital, your doing what EVERYONE else in the hard drive industry is doing! Sucking!

Re:16 mb cache eh?

[v1]

but there is no reason why a hard drive can't have multiple read/write heads to access different sectors of the disk as the same time

It's called "seek time". The more massive (heavy... more parts, more heads) the read arm has, the slower the seek time is. The heads have to travel a shorter distance, but it requires more time to move them that distance. You're robbing Peter to pay Paul.

Though the idea of a single drive striping, say across platters, is a nice idea. Really even that could be done with one platter, since each platter has two surfaces. (top/bottom) Though that leaves many points of single failure still... onboard controller card, stepper motor, arm mechanics, any of which can fail and cause total data loss. About 1/2 the drive failures I've seen involve the drive no longer being able to spin up or stay spun up, so I don't think this can be considered anywhere near as failure-proof as say, a two drive mirror.

Submitter is kind of clueless?

[hklingon]

Ok. All slashvertisement comments aside, I get as excited about 'teh new hotness' in drives as much as the next person. But this is SO poorly submitted. 300MB/sec? PLEASE. You MIGHT get 70% of that speed doing a transfer from that 16mb buffer to the controller, but that is just misleading. Without even reading, I'm guessing they're talking about 3Gb/sec SATA-II. Woo. So that is wrong. "Interface Speed" is what you wanted to say there. Not "Transfer Rate".

What about "WD Characterizes this as the highest performance section of the desktop market." Wrong again. Helooo??? Raptor??

I mean. Talk about something cool, at least. New TCQ optimizations? Read-before-write? 24/7 100% duty cycle?

SR is a decent place to check out reviews and benchmarks. Do your homework! Astroturf like this only spreads confusion and disinformation.

I got a 15k RPM SCSI drive from hypermicro. It is a seagate, 73gb. It was only about $250 with an adaptec controller (which wasn't a whole lot more than a WD74 gb raptor at the time). At the beginning of the disk, it has over a 90Mbyte xfer rate on a 160mbyte/sec interface, which totally crushes all this other crap. My drive is (was?) the leading drive on non-raid configurations on hdtach's website, even against the 400gb SATA WD behemoth. 2x36gb raptors are about the same speed as one decent 15k RPM scsi disk.

I haven't really looked, but I would guess the drive in the post is what.. neighborhood of 60mbyte/sec? 70? Meh. Meh I say. We didn't even talk about I/Os/sec. between 7200 rpm, 10k RPM and 15k RPM.

The idea of an article like this on slashdot is not bad. It is just that this article is misleading and/or wrong and isn't really news at all. And so on and so forth.

Silly Question...

[Manip]

So it runs at 300 MB/sec; but can you get 300 MB from it? .. I mean throwing out general performance numbers like that is completely meaningless.. How much of that data is in the cache? Does it before at 300 MB/sec for more than 16 MB? What if I do a number of seeks, how long does it take then?

Point is you can't just create numbers and throw them out... The fairest way to do it is to compare a few similar drives using identical testing software that reflects real life read/writes on a disk over a period of time.

I would also like to see advances made in drive redundancy; far more so than speed. Why is it when I have four or five platers in a drive, that any one failure can cause a 100% data loss? Shouldn't the data loss be limited to just that plater or read head? ... Perhaps a little R&D in that area, I know I'd pay more for data security.

Sustained transfer rate?

[miffo.swe]

I dont care about spurious theoretical cache transfer rates. What i care about is the sustained transfer rate and the ability to do more than one thing at a time. Come to think of it i think i really hate HDs. When o when will we have solid state long time memory in our computers without moving parts?

Nice but...

[snevig]

Storage Review has the Hitachi 7K500 as the best desktop performer out there right now.

Their review of the WD2500KS compares it to the Hitachi 7K400 and the WD clearly loses out.

The 7K500 is compared to the 7K400 in its review and the next-gen performance boost is quite clear.

Just Cheaper, Please

[Doc Ruby]

I'm pretty happy with drive performance. All I really want is lower per-byte prices. A RAID means drives can deliver data in parallel for faster data transfer; multiple RAIDs on a SAN or a PCI even faster. I want all their R&D going into making it cheaper. HDs right now cost $0.31:GB for 250GB drives. When that's down below $0.05:GB, I'll be interested in hearing about faster transfers on individual drives.

Re:Doesn't do any good if

[billcopc]

Statistical manufacturing defects aside, I'd say the #1 reason for a failed drive is a cheap system. That 200$ PC uses a 7$ power supply and no fans except the CPU. Not all 300watt power supplies are created equal.

These hard drives need juice and they need cooling. I have been a Maxtor nut forever because they run faster than any WD or Seagate, but they run hot. The average moron with their PC in a desk drawer will kill one of them in the first month. I've been running my raid-0 for two years now without a hitch. The trick ? I always stick a nice big 120mm fan right across the hard drives, blowing around them. And believe me, they don't sit idle all day. I've got 6-way Raid-0 for a reason; I thrash them harder than your average SQL server.

Now I'm citing Maxtor because they tend to run the hottest (performance driven), but this applies to all hard drives. If you let heat stagnate, you will kill any electronic device, it is merely a matter of time. Take care of your gear and it will last a long time.

Re:Doesn't do any good if

[WuphonsReach]

Yep yep yep.

I learned that lesson after killing a 7200rpm SATA drive. (Actually, I learned that lesson back in 1998 with SCSI drives, but I was being lax.)

That's why I like things like the newer PC cases that put the drives sideways and stick a 120mm fan to pull air over them. (Antec Sonata, Antec p160, etc.)

The other key bits in my toolkit are bay coolers. One lets you put up to (3) 3.5" drives into (2) 5.25" bays (try MWave for these), the other is a "4 in 3" bay design (CoolerMaster). If you don't pack them full (e.g. only put 2 drives in the 3:2 unit, or 3 drives in the 4in3 unit), the 80mm/120mm fans on the front keep even hot 7200rpm drives cool to the touch. Plus, the 80/120mm fans are *quiet* when compared to the normal 2x40mm fans used on regular 5.25" bay coolers.

As for power... I used to have a (6) drive ATA RAID that would constantly drop a disk every few weeks. Nothing physically wrong with the disks, the array would simply drop the drive and start rebuilding with the hot spare. Upgrading to a better UPS along with a better case/powersupply fixed the issue for good. (The array stopped dropping drives.)

Re:Doesn't do any good if

[WuphonsReach]

Some people get hurt and then complain in real life and on the net and then replace that drive with some other manufacterer, get lucky with that drive and proclaim how great that company is.

And the rest of us plan for failure by using RAID in addition to backups (and system images). I hate running systems without RAID, because I *know* that eventually that drive is going to fail at the worst possible moment.

(I probably have close to a dozen IBM "Deathstars" (the 72GB models that everyone hated) that are still chugging right along. I've probably replaced *2* of them in the past few years and those drives failed early in the cycle and those were likely killed by heat.)