Hard Drives Made for RAID Use

An anonymous reader writes "Hard drive giant Western Digital recently released a very interesting product, hard drives designed to work in a RAID. The Caviar RE SATA 320 GB is an enterprise level drive without native command queueing and uses an SATA interface. In works better in RAID than other drives because of features like its time-limited error recovery and 32-bit CRC error checking, so it is an option when previously only SCSI drives would be considered."

Comments

Slashdot: Stories Made For Ad Use

[Anonymous Coward]

Sheesh, this is a VERY thinly disguised ad. Here's a direct link to NewEgg $169. Has the same details as this "story."

Re:Slashdot: Stories Made For Ad Use

[andrewman327]

Does anyone have any benchmarks to back up this claim? This seems very vague.

Re:Slashdot: Stories Made For Ad Use

[fimbulvetr]

On the newegg link they list the MTBF as 1 million hours. Google tells me that that is about 114 years. How can it have such high mtbf? Is that newegg just not having correct data or is there something special about these drives (or are they designed to be "used" less)?

Re:Slashdot: Stories Made For Ad Use

[ptbarnett]

Is that newegg just not having correct data or is there something special about these drives (or are they designed to be "used" less)?

It's not an error by NewEgg. Follow the link to the manufacturer's site, and you'll see the same specification:

http://www.wdc.com/en/products/Products.asp?DriveI D=114

Re:Slashdot: Stories Made For Ad Use

[cperciva]

On the newegg link they list the MTBF as 1 million hours. Google tells me that that is about 114 years. How can it have such high mtbf?

MTBF is defined as [short time period] * [number of drives tested] / [number of drives which failed within that time period]. An MTBF of 114 years doesn't mean that half of the drives will survive for 114 years without a failure; it means that if you run 114 drives for a year, you should expect to have 1 failure.

A more intuitive way of conveying the same information is to say that the drives have an expected failure rate of no more than 1E-6 per hour.

Re:Slashdot: Stories Made For Ad Use

[Rasta Prefect]

On the newegg link they list the MTBF as 1 million hours. Google tells me that that is about 114 years. How can it have such high mtbf? Is that newegg just not having correct data or is there something special about these drives (or are they designed to be "used" less)?

Easy: You, like most people, don't know what MTBF means. MTBF is only meaningful in context with the expected lifespan of the device. This is probably somewhere in the neighborhood of 5 years, or about 43,800 hours. Essentially, what the manufacturer is saying is "Based on some data, we estimate that if you run x number of these drives, the average time between failures will be 1,000,000/x hours, up until the expected lifespan of the drive, at which point all bets are off"

For computer hardware this is always some sort of extrapolated estimate, since they have of course not actually been testing the drive for it's expected lifespan, or it would be obsolete by the time they released it.

Re:Slashdot: Stories Made For Ad Use

[tomhudson]

Theoretically, any cheap drive used in a raid will experience less wear per gig of RAID data storage, since it is only storing a portion of the data. It's a cheat. Also, MTBF is a theoretical extrapolation from failure time of individual components. In the hard disk industry, its relation to reality is about the same as Harry Potters'. But we should be used to that, just like a megabyte ain't a megabyte when they calculate capacities.

Its like this quote from the article:

In (sic) works better in RAID than other drives because of features like its time-limited error recovery and 32-bit CRC error checking, so it is an option when previously only SCSI drives would be considered."

It's all bullshit. Sure, it might be better than another drive for use in a raid, but its not like people couldn't consider IDE drives in the past, and that this is some miracle cure.

Just look at what RAID means - Redundant Array of Inexpensive Disks. Lots of people use cheapie IDE hard disks in RAID setups. We've got a 4-drive terrabyte raid. Why would we consider expensive drives when the whole idea is to use cheap drives in a redundant array?

Fuck the marketing departments. And fuck the PHBs who make their buying decisions based on them. Oh, right, the PHBs *ARE* getting fucked by the marketing departments. Sorry lads, carry on.

Re:Slashdot: Stories Made For Ad Use

[fimbulvetr]

Thanks.
I was actually wondering when it went up to 1m, it has been a while since I've spec'd out drives, but I seem to remember that 250k-500k was the norm.
It's interesting that the mtbf has doubled (or even quadrupled) since the last time I remember.

Re:Slashdot: Stories Made For Ad Use

[Mozk]

Since when does saying that you agree get you modded 5, Insightful?

Re:Slashdot: Stories Made For Ad Use

[raydobbs]

RAID - Redundancy Across Independent Disks

Re:Slashdot: Stories Made For Ad Use

[Anonymous Coward]

So, what was the MTBF of the IBM Deathstar/Deskstar drives? Shows just how useless MTBF numbers are.

Re:Slashdot: Stories Made For Ad Use

[whoever57]

An MTBF of 114 years doesn't mean that half of the drives will survive for 114 years without a failure; it means that if you run 114 drives for a year, you should expect to have 1 failure.

That is a good explanation. Many people confuse MTBF with lifetime.

Most products (and especially electronics) have a failure rate that when plotted over time looks like a bathtub. There is a high initial failure rate (infant mortality) that drops over time to a base rate (the random failure rate described by MTBF), this low failure rate continues until one reaches the end of useful life of the product, when the failure rate rises once again as age and wear effects cause the device to fail.

Note that most extended warranties are designed by the seller to kick in after the early failure rate has droped, but expire before the end-of-life failures.

Re:Slashdot: Stories Made For Ad Use

[sakusha]

Correct. MTBF is not designed as an index of reliability for any one specific drive in use. It is designed as an index for manufacturers and repair facilities to estimate how many spares are required per year for any widescale deployment. So if you have 114 drives deployed in your enterprise, you would need to stock 1 spare drive to replace the estimated failures in one year.

Re:Slashdot: Stories Made For Ad Use

[Jozer99]

They don't run a drive for a million hours, they run 100 drives for 1000 hours, and one breaks. Drives are getting better and better. I don't see how this is unusual, in my expierence, if a drive works for the first two weeks it is installed, it usually lasts for decades if it is properly stored (not running in dust, properly cooled, ect... That is of course, unless it is a Maxtor, in which case, get your fire extinguisher.

Re:Slashdot: Stories Made For Ad Use

[KillShill]

thank you for that information. i assumed incorrectly that at least this one spec from the manufacturer wasn't a complete and utter sham. now i shall remain ever more vigilant.

gee, one wonders if anything the manufacturers of products say is true.

LCD monitor manufacturers lie just about everything on the specs. Hard Drive manufacturers lie about an enormous amount about their products.
software vendors lie a ton about their products and "fitness or lack of for a particular purpose" (then why the hell are you allowed to sell junk if you don't guarantee at least a minimum set of suitability).

and the list goes on and on.

now when i see 1 Million hours of MTBF i'll know it's just another piece of crap marketing/advertising lie.

yep, more "honest commerce" at work.

Re:Slashdot: Stories Made For Ad Use

[Anonymous Coward]

When you whore with a 'here goes my karma' sentence. You see, the mods like to think they are more intelligent than the rest. More insightful. More fair. In reality, they are a bunch of dumb fucking lemmings. So they see this post with the stupid little phrase, and decide to mod up. The rest of the post could be "but I think brown stripes make my underwear more fashionable". The slashots would nod their and say "fight the man!" and mod up.

Re:Slashdot: Stories Made For Ad Use

[eh2o]

It also says "24/7 reliability" which I think means "100% duty cycle" so ostensibly they are not designed to be used less (as most ide desktop type drives are).

IIRC Seagate is the only other company to offer a 5 year warranty on ide type drives (also subject to proper use -- no desktop drives in servers).

Due to fluid dynamic bearings, better motors and other former SCSI-only technology the reliability of ide type drives has gone up a lot in recent times (thank god).

Re:Slashdot: Stories Made For Ad Use

[Hack Jandy]

Cheaper at ZipZoomFly:

http://labs.anandtech.com/search.php?q=WD3200SD

HJ

Re:Slashdot: Stories Made For Ad Use

[petermgreen]

i've seen "redundant array of inexpensive disks" (which i belive was the original) and "redundant array of independent disks" but never the one you mention. Care to cite a source?

Re:Slashdot: Stories Made For Ad Use

[Anonymous Coward]

Theoretically, any cheap drive used in a raid will experience less wear per gig of RAID data storage, since it is only storing a portion of the data.

No, it will receive MORE wear per gig. If I write 1 gig of data to a raid, more than 1 gig of data will be written to the drives, since the redundancy data must be written as well as the regular data.

Re:Slashdot: Stories Made For Ad Use

[Danga]

Theoretically, any cheap drive used in a raid will experience less wear per gig of RAID data storage, since it is only storing a portion of the data.

That is not the case in RAID 1 where the data is mirrored. It may result in less reads per disk compared to a one disk system but the data is definately not portioned out across a mirrored disk RAID setup. I do agree that it's not a bad idea to use cheap disks ESPECIALLY if the data is mirrored or at least backed up regularly if using RAID 0. I wouldn't spend the extra money on these "special" HDD's.

Re:Slashdot: Stories Made For Ad Use

[Bachus9000]

Care to more specifically define "proper use" for Seagate's drives? I've got a few of their drives in a Mythtv/file server box... :)

Re:Slashdot: Stories Made For Ad Use

[dtfinch]

MTBF is the failure rate of drives that are neither defective or worn out. If I had to guess, it's an estimate of the lowest point on the failure rate bathtub curve, maybe around 2 years into the life of the drive. Defective drives usually fail within 1-2 years, and the rest start to wear out after 3-5 years.

Re:Slashdot: Stories Made For Ad Use

[arodland]

Or you could stop being an idiot, realize that it's a perfectly useful number that businesses find important, regardless of the fact that you were sadly deluded as to what it meant, and stop whining about stupid shit on slashdot.

Re:Slashdot: Stories Made For Ad Use

[tomhudson]

No. If I write 1 gig of data to a 5-disk raid, (4 disks for data, 1 for parity), then only 1/4 gig gets written to each disk, including 1/4 gig to the parity disk ...

If I do this for a 3-disk raid, it still only ends up with less than half as much data ( 41-2/3%) on each disk, even including the parity data that is distributed among the partitions.

But yeah, even simple striping will also reduce wear, though it won't give you redundancy. However, anything that reduces wear per drive is better for your data than nothing at all :-)

Re:Slashdot: Stories Made For Ad Use

[Tassach]

When RAID was first deployed, the I was originally "inexpensive". "Inexpensive" in this context meaning "cheaper than the $100,000 washing machine sized disks you have now". Think old-school VAXen and PDP-11s. "Inexpensive" transformed to "independent" when the industry moved away from big iron to modern server hardware using commodity SCSI disks.

Re:Slashdot: Stories Made For Ad Use

[Anonymous Coward]

I work for a company where we have sold hundreds of thousands of diskdrives. We also have the ability to remotely monitor the health and replacement rate of a huge percentage of those drives and see how they, as a population, perform in the field.

Claiming 1,000,000 MTBF drives is a joke. Try 500,000-600,000..maybe 700,000 in a very well run and environmentally controlled datacenter. Substantially less in not-so very well run datacenters.

Re:Slashdot: Stories Made For Ad Use

[tomhudson]

This is true for mirroring (which I obviously wasn't referring to when I said that only a portio of the data is stored on each disk); however, in applications where there are few writes and lots of reads, it still ends up saving wear and tear, because the controller can interleave reads from both disks, giving 2x the throughput, as well as only half the wear per drive.

Cheapie disks are the way to go. Spend the money you save on a second box with a complete backup as a hot spare.

Re:Slashdot: Stories Made For Ad Use

[eh2o]

Check the specs on the drive to see if the drive is rated for 100% duty cycle or not. If not then to maximize drive lifespan the drive should be powered down (or sleeping) for about 50% of the time (usually when you are sleeping, too). If you use one of these drives in a 100% uptime environment (e.g. web or e-mail server) it will usually fail prematurely (within about 2 years in my experience). The engineering of the motors, actuators and bearings simply cannot withstand abuse (probably a combination of insufficient heat management and less precise tolerances in the manufacturing of parts and/or cheaper technology).

In the fine print of the Seagate warranty they mention that they won't cover this type of failure unless the drive is rated for server use.

Re:Slashdot: Stories Made For Ad Use

[Anonymous Coward]

I visited the parent poster's link and verified that the spec for MTBF is the same.

Consider this informative and moderate it appropriately.

Re:Slashdot: Stories Made For Ad Use

[Anonymous Coward]

On the newegg link they list the MTBF as 1 million hours. Google tells me that that is about 114 years. How can it have such high mtbf? Is that newegg just not having correct data or is there something special about these drives (or are they designed to be "used" less)?
 
Try more like 10 hours. I ordered 10 of these drives for a raid storage cabinet at my office. Within a week, 6 of the 10 drives failed. Some clicked. Some spun up but then spun right back down. I tried to see what the error codes were using the western digital harddrive diag toolset. One came back with an unrecoverable error, the others weren't readable.
 
These drives were shipped to me by my distributor from two different locations (because of stock issues), so unfortunately, its not because the entire box fell 30 feet from the plane or any such nonsense. I have tons (160+) of the WD250 raid edition drives in use, and they work great. I think I've replaced two of them over the last year. These 320s though proved to be complete crap.

Re:Slashdot: Stories Made For Ad Use

[Diag]

MTBF isn't a particularly useful spec for home users. But it's certainly relevant to businesses who purchase disk arrays that can contain hundreds or even thousands of disks.

Re:Slashdot: Stories Made For Ad Use

[ahaning]

5-disk raid, (4 disks for data, 1 for parity)

But I thought RAID5 spread the parity across at least 3 disks.

I think you described RAID4.

See the Wikipedia article on RAID4.

Re:Slashdot: Stories Made For Ad Use

[ahaning]

Whoops. On rereading your post, you didn't specify the level of RAID you were talking about, just the number of disks.

Carry on.

Re:Slashdot: Stories Made For Ad Use

[KillShill]

perhaps but it's very misleading... hence it's marketing/advertising.

just like how they misrepresent the size of the hard drives using the non-standard method of measuring. and also how they use different ways of determining the seek time, latency etc.

frankly people buy drives anyway... why lie about what they're really capable of.

Re:Slashdot: Stories Made For Ad Use

[drgonzo59]

You forgot the good ol' IBM "Deathstars". I got one of those that lasted a good 2 years then "click..click..dead!" taking with it some family photos and digital copies of some documents, as well as a good couple of weeks of my project work.

Re:Slashdot: Stories Made For Ad Use

[Jozer99]

Ya, never had much occasion to use the IBM drives, except one in an iBook (!!!), which died after 3 years, I took it out, tested it in another machine as verified completely dead, put it in a box, two years later take it out and it is good as new no errors. I still use it for movies ect...

Re:Slashdot: Stories Made For Ad Use

[raydobbs]

Your second one there is correct - I just remember the diatribe the 'Inexpensive' part gave a teacher of mine - mostly due to the fact that the expense of the disks had nothing to do with the actual make up of why you'd make a bunch of them perform as one larger one with better data security than is possible in a single drive.

Re:Slashdot: Stories Made For Ad Use

[iamplasma]

Well, at least they're better than the metamoderators. At least when a mod sticks a retarded moderation on a post, other more sane people can mod it the other way. However, I recently got a metamoderation on one post I had modded "flamebait", because all it said was something along the lines of "you should eat flaming shit while smoking a fag pipe and taking it up the ass" (you get the idea). Apparently, that was unfair. Riiiight, my loathing for people who say "I vote all negative moderation unfair" just went up yet another notch. When do we get meta-meta-moderation?

(And yes, I'm making a -1, Offtopic post now, so shoot me, I won't vote your moderation "unfair" for it)

Re:Slashdot: Stories Made For Ad Use

[Anonymous Coward]

What you describe is the expected lifetime! Then you say people confuse it with lifetime...

That, my friend, is what's confusing...

Re:Slashdot: Stories Made For Ad Use

[tomhudson]

Hey, its all good :-)

YOu can have a 9-disk raid if you want,or even a 50-disk raid of raids (ouch). The only problem is, as the number of disks goes up, the likelyhood of near-simultaneous failure of more than one disk goes up, too.

Sort of like the bad old days of tube computers. If you had 4000 tubes, and they had an average lifespan of 8000 hours, you'd be lucky, on average, to get 2 hours of uptime before one burned out. Then , since you had to keep the tubes "lit" so you could find the dead one, by the time you replaced it you were lucky to get 15 minutes of runtime before another one blew.

Its like the shuttle engines. They spend so much time testing each one that by the time it's ready for launch, they've already used up 90% of their rated burn time. Google for Feinman's report for more info on NASAs bs flight practices. Seems NASA and Microsoft have some similarities in management.

This is ridiculous!

[Cerdic]

If they would stop eating around the hard drives, leaving crumbs in them, we wouldn't need to use Raid to take care of the cockroaches in them. Ugh.

Re:This is ridiculous!

[Compaq_Hater]

how the hell do you get crumbs inside of a Sealed unit such as a Hard Disk ?.

you by chance are not one of those Idiots who thinks that a Tower or computer Case is a Hard Drive are you ?.

If you answer Yes to the Latter You are a MORON !

CH

Re:This is ridiculous!

[FunkySquid]

It was a joke. http://en.wikipedia.org/wiki/Raid_(insecticide)

Re:This is ridiculous!

[MrKibkibs]

RAID: Kills bugs dead. Can't do anything for your sense of humour though.

No NCQ?

[gbjbaanb]

Interesting that they don't have NCQ, whereas SCSI drives generally do (well, called TCQ on SCSI IIRC)

Is this just marketing speak, has it truly included scsi features, or could it actually be better performing than SCSI in a RAID array?

Re:No NCQ?

[Baddas]

NCQ = Native Command Queuing... as said in the article.

Re:No NCQ?

[Baddas]

Ooops, I'm an idiot, disregard that. Damn advertising writers and their extensive use of negative statements. :)

Re:No NCQ?

[rsborg]

Is this just marketing speak, has it truly included scsi features, or could it actually be better performing than SCSI in a RAID array?

In snort, without NCQ, SATA drives are going to be slower than SCSI. The other two features probably just offset/mitigate the speed differences, but I would probably hold out for something that has NCQ (or just go SCSI) if I were building a RAID today.

Re:No NCQ?

[keltor]

I think that all SCSI RAID controllers disable it on the drives as the controller takes care of all queueing. Remember most of the SATA drives now have NCQ. WD chose to specifically disable this as their regular Caviar SE drives have queueing.

Re:No NCQ?

[Anonymous Coward]

SATA RAID products have been around for a while now. So I don't understand what it means to say we finally have a SATA drive suitable for use in a RAID array. Have all SATA (how about SATAII?) arrays prior to this been complete garbage, because there is not suitable hard drive? Something doesn't smell right.

Re:No NCQ?

[GigsVT]

The summary is wrong, it (the RE2) does have NCQ.

See a real review like TechReport instead of that amateur crap they posted.

Typo!

[Anonymous Coward]

"In works better in RAID..."

You should change "In" to "It"

Thank you very much.

Re:Typo!

[Anonymous Coward]

It works better it RAID... No, that doesn't work either.

Re:Typo!

[Basje]

You case insensitive clod!

About time

[Tuor]

While I've been a proponent of SCSI for a long time -- Apple really was thinking ahead when it had it in Macs all those years -- it has been getting thread-worn. Ultra-wide-tall-double-hex-SCSI is just getting to be too much!

SATA is the right technology, especially for controllers since each channel is dedicated. The only alternative is Firewire, and there are no native controller drives.

Re:About time

[mgpeter]

This might be a good alternative to SCSI drives, except it is only 7200rpm !?!

Why would Western Digital market THIS drive for RAID configs when they have 10K rpm SATA drives (Raptors) they could have used instead ?

Re:About time

[Anonymous Coward]

Because the raptors stop at 74Gb and this is 320..?

Re:About time

[$RANDOMLUSER]

> ...especially for controllers since each channel is dedicated...

I generally tend to agree with that, but as a guy running 8 200GB SATA drives on four controllers, I can tell you that the PCI bus gets saturated _way_ too quickly for my tastes.

Re:About time

[Glonoinha]

RTFA - they used a different type of encoding on these drives in order to implement the 'time-limited error recovery.' The problem is that the encoding is done on three-vector bi-furate substrate instead of the two-vector bi-furate substrate used in the Raptors, and the 3V stuff can't handle speeds of the 10k RPM (the lateral acceleration at 10k RPM is significantly more than at 7,200 RPM, and the 3V stuff is taller than the 2V stuff - hence the problem.)

Re:About time

[Anonymous Coward]

The only alternative is Firewire, [...]

No, It isn't. Fiber Channel is the alternative. Additionally, beeing on dedicated channels is only acceptable if you stick everything in one case, and it really is only any good in SATA's case because parallel IDE did such a bad job of handling 2 drives on one wire. With proper arbitration (like SCSI or FC/AL have), there really isn't much of a problem with having multiple drives on one bus.

Re:About time

[Ilgaz]

Well, serial attached scsi started to ship

http://www.adaptec.com/sas/index.html?source=home_ story1a_SAS_technology_home

Pro level already moving but I suspect it will be OK for home with enterprise features it offers.

I checked a bit you know ;)

Re:About time

[sumdumass]

8 drive on four controlers.

You could get around that if you were to use a Adaptec Serial ATA RAID 2810SA with 8 ports or a more expensive Adaptec Serial ATA RAID 21610SA with 16 ports.

You might look at the price and say too expensive but the speed and availible configuration should make up for it. Besides i got might for around $425 wich is less then thier suggested price. Also both these cards can use the waisted space from mismatched drive sizes as well run multiple raid volumes one each drive. What i like the most is the hotswap and hotspare were you could just leave a blank drive in and if one other drive failes it automaticaly recovers with the spare and you can replace the bad drive without rebooting. Another thing i like about the card is that it is a full controler and not one of these host based things. Your computer will just see it as a harddrive(s) without any special drivers. You can even access them from DOS, most linux kernels, as well as windows 95 and 3.11 (note the drives had to be small for 3.1 and 95 to see correctly).

BTW, i don't work for adaptec or sell thier stuff. I'm just impressed with a product that finaly took alot of frustration away that has been associated with cheaper IDE and SATA ad-on cards. I'm sure there are better solutions availible. this is just one that i have found. Most of the cheaper (under $100) IDE,SATA,or raid controlers i have found use the system for thier existance. This is why you need a special driver in windows or linux to use it corectly. the extra cominucations here could be somethign saturating you pci bus (or helping it saturate)

Re:About time

[Fweeky]

Quite. 32bit 33MHz PCI (especially shared among on-board stuff *and* multiple card slots) is amazingly feeble these days, so consumer-level PCI Express comes not a minute too soon. Of course if you can afford and appreciate 8 200G drives you can probably also afford and appreciate a half-decent workstation/server board with PCI-X, but even a pair of modern drives can completely saturate the bus, and if you're into file sharing over GigE even one drive is way too much.

For that matter even sharing /dev/zero over GigE on PCI is.. disappointing.

Re:About time

[timeOday]

Why would Western Digital market THIS drive for RAID configs when they have 10K rpm SATA drives (Raptors) they could have used instead ?

AFAIK the Raptor is still stuck at 75GB, no? That's getting downright pathetic.

Individually, each of these new drives is slightly slower than a Raptor. But the cheap price for high capacity would allow liberal use of Raid-1. A pair of these in Raid-1 should destroy a single Raptor in every read benchmark.

Re:About time

[Anonymous Coward]

Sharing /dev/zero goes a lot faster if you turn on compression.

Re:About time

[Fweeky]

They're AAC devices, I believe. If their SCSI models are anything to go by, I'd say.. avoid like the plague. Between buggy firmware, overheating hardware, crappy drivers and even crappier management tools, you're probably better off spending that money on a motherboard with a decent bus (which you'll need to drive any non-trivial RAID card anyway) and just using software RAID on a known-good controller. This is based on several years experience with a number of 2120S cards and various systems and OS's.

Seriously, this is the company which sells a £50 dual port SATA card based on the same SiI chipset you find in £5 cards; that same chipset widely regarded as being one of the buggiest controllers in common use today. I shudder to think what they fit behind their proper SATA RAID controllers.

Re:About time

[Fweeky]

Well, yeah, the sizes of higher end drives tend to be pretty poor because they use smaller, less dense platters to aid seek times (less distance for the head to travel, less time waiting for the head to settle on a track). These drives quote latencies of ~9ms; Raptors and SCSI drives start at around 4.5ms and get as low as 3.

In a server environment where you're limited by the number of drives you can fit in your icke pizzabox (and power and cool effectively) you can't always just say "let's throw twice as many drives at the problem" if it's raw IO's/sec that you're worried about rather than space or serial transfer rate.

Re:About time

[sumdumass]

I cannot seem to find were it says that is an AAC. here are some specs for it. I'm not exactly sure what AAC is anyways. I just know these are better then the $70 cards that needs the aray rebuilt for each operating system installed. With the 2810sa, i was able to access it from windows, linux installed or not (old versions as well as new) and even dos without a special driver or software program installed.

Seriously, this is the company which sells a £50 dual port SATA card based on the same SiI chipset you find in £5 cards; that same chipset widely regarded as being one of the buggiest controllers in common use today. I shudder to think what they fit behind their proper SATA RAID controllers.

Yep and thats why i sugested a real hardware raid controler. Its powered by a Intel 80303 100 MHz and is basicaly self contained. The product brosure says it is for midrange servers and workgroups.

I'm interested in if you could recomend a better card for the same price range.

Re:About time

[Fweeky]

AAC is "Adaptec AdvancedRAID Controller", and yeah, I just checked and that model is indeed one.

It is indeed a real hardware controller, but it's quite likely they have a bog standard SATA controller behind the 80303; I think they have a AIC79XX behind the SCSI models, but I dunno, I've never had a good look at the physical card.

They're not *awful*; our master database has ran one for years with few problems (FreeBSD; uptime currently 218 days, with ~5.2 billion queries), the main one being disks randomly popping off the array for no apparant reason, which could be drive firmware related. Still, we had enough problems with the controller occasionally stopping processing IO's for no apparant reason (mainly on Linux) to think that maybe they're not the most robust devices on the planet. Google should confirm that these aren't the most problem-free controllers about.

Better card? Well, LSI seem to know what they're doing, but after finding the latest driver doesn't want to talk to our card (maybe because we have >8G of memory) I'm dubious there too, although their hearts seem to be in the right place with plenty of community interaction and support. 3Ware have a lot of people raving about their SATA cards but I have no experience with them.

So, er, yeah, I'd be interested in recommendations too :)

(Currently fighting a 2120S; "No partitionable media were found." says debian-installer. I'll be glad when we're back to software RAID.)

Re:About time

[Huge Pi Removal]

I built an archive server a few weeks back using one of those 16-channel Adaptecs, on a nice Tyan motherboard with PCI-X. Works a treat, I'd definitely recommend it.

Re:About time

[sumdumass]

I'm glad to see that they are not *awful*. In fact they are better then what i was working with. I didn't install their software that comes with it so maybe i'm not going to see too many issues with it.

as for the 2120s, I have a couple of these drom some old novell servers and i too am experiencing some issues. I read somewere that said you couldn't boot to them if the IDE ports were enabled on the mainboard. I don't know if that is giving your issues or not. (somethign about the bios overlay wouldn't instal with the IDE ports enabled and you could only use them after the OS was installed.) I run (6) SCSI cdrom drives from one and one drive keeps loosing the directory listing. I thought it was because i originaly installed withn a cd writer wich was later removed. The default instal location comes back to the third drive in the chain wich is were the SCSI emulation device placed the burner when i did my instal and happens to be the cdrom giving issues. Maybe my problem isn't with changing drives out but the 2120 card itself. I'm about to reload it anyways, if your interested, i could let you in on anythign i discover.

Thanks for your insight into this. I'm not a raid expert by any means. I'm always open to learing from others experiences and i apreaciate the time to took to respond about the AAC and giving me another product to look at. I plan on building a couple more file servers in the near future and will be looking to check some of these LSI and 3Ware products out. Right now with the 2810 and a gig of memory, AMD64 processor, and a trimed down mandrake version, this server is the fastest on our network (access time as well as speed). Aplications that once took 2 minutes to open now open in less then 20 seconds. (i know that extreamly lng but i couldn't do anythign about it myself unntil one of the servers drives failed taking everythign with it. And the tape backup took almost two weeks to recover).

Re:About time

[Fweeky]

The only software of theirs we've used other than the drivers bundled with Linux/FreeBSD have been aaccli; a management interface is pretty important with a RAID card :)

The smallest system we've ran them on is a dual 2.8GHz Xeon with 4G of memory; we're currently testing one in a dual 265 (dual dual core 1.8GHz Opteron) with 16G of memory and 8 15kRPM SCSI drives in RAID10, and while it works (albeit throwing Machine Check Exceptions under IO load), performance is merely "sufficient"; Linux md raid10 is generally faster, more flexible, and seems less likely to fail horribly during normal operation, although a proper RAID card with a battery backup module would probably manage power failures better.

It's certainly possible to be lucky with these cards, probably down to picking the right combination of firmware, driver versions, drives, motherboard, IO load, and orientation with the nearest layline. It's a lot more hit-and-miss than I'd like, though; for something that's supposed to increase reliability and uptime, we've probably had more downtime from card failures than drive failures. YMMV.

Intel make MegaRAID cards based on the same design LSI use, btw; may be worth a look if you're not looking at systems with tonnes of memory. They do PCI Express versions too, making them one of the first PCI-E IO cards I've seen.

How does a lack of NCQ help?

[Nom du Keyboard]

How does the lack of Native Command Queuing improve RAID performance? Generally I thought NCQ improved all drive's performance, and TFA says that NCQ is normally part of Enterprise High-Performance.

earth to 11 year old kid

[Anonymous Coward]

You can't sell your oh so cool hardware review site for millions of dollars and retire at age 12 just because Cowboy Neal posts your article on slashdot!

p.s. Pay attention in English class.

Summary of article:

The Good (+)
- Very good performance
- Looks cool (for a hard drive)
- Optimized for RAID use

The Bad (-)
- High initial investment

Re:earth to 11 year old kid

[alc6379]

Mod me offtopic, or whatever, but this has to be the most dumb-ass review I've ever read. It's a drive meant for RAID use, as in RAID 5 or RAID 1, in servers, where data integrity is very important. But what does this guy do?

...he puts it through the paces of a desktop hard drive. Where's the test of how it could run under mySQL? It's been replaced by a comment about how you can never have too much space "in this age of DVD-burning, file-sharing, and 40 GB MP3 players." Who the fuck cares about that on a server?

Where's the review of how well it facilitates serving pages through Apache? Oh, that's replaced by "Look how neat the drive looks!"

...Nope. This FA was a waste of time, not just for the reader, but for the author, and for Western Digital to have even sent the drives to this guy. He should go back to playing UT2k4OMFGBF2, and find someone who actually knows something about industry usage patterns on hard drives like this to write a thoughful review.

Re:earth to 11 year old kid

[Anonymous Coward]

It's a drive meant for RAID use, as in RAID 5 or RAID 1, in servers, where data integrity is very important. [...] Where's the test of how it could run under mySQL?

If data integrity is very important, you're not running a buggy toy like MySQL.

Re:earth to 11 year old kid

[alc6379]

If data integrity is very important, you're not running a buggy toy like MySQL.

Ok, fine:

Where's the test of how it could run under insert database name of choice here ?

Happy now?

SATA version may be new, but features are not new

[ptbarnett]

Western Digital has been selling an EIDE version with this feature set for a while:

http://www.wdc.com/en/products/Products.asp?DriveI D=92

I bought one to replace what I thought was a bad drive in a RAID configuration about a year ago.

TechReport

[JohnnyBigodes]

Proper TechReport's review here.

Go read. Now!

Re:TechReport

[andrewman327]

Why didn't the original article link to the TechReport review?

To be fair, it looks like the drive really is impressive. From the review: In the world of enterprise-ready 7,200-RPM Serial ATA drives, the Caviar RE2 has few competitors and no equals.

native command queueing

[garat]

Here's an interesting quote from Tom's Hardware:

"In sum, we must state that all Command Queuing enabled drives have an advantage over those that do not support this feature. At the same time, CPU load is also slightly higher when Command Queuing technologies are used. However, considering the performance of today's processors, the additional CPU load is a marginal factor."

Basically, you put some load on the processor for increased disk performance... Why not include it?

Re:native command queueing

[sumdumass]

Probably because of the mirrage of software raid cards that use the system processors and memory. Are we talking about an extra load on the system's processor, The raid controlers or the little IDE's processor? (yes there is a small proccessor on most IDE or SATA drives that do LBA)

Many of the cheaper level controlers do this exact thing but apear to be a hardware controler. OTOH, i'm not sure if a true hardware controler would be able to take advantage of it either. Are there current SATA raid controlers even able to use Command Queuing? If not how long until it is possible. It may be a feature that couldn't be used at this stage but later revisions might.

Re:native command queueing

[jandrese]

Because that functionality is supposed to be handled by the RAID controller, and if the drive has it enabled you can end up losing performance as the two Command Queue Controllers fight each other. Theoretically you can turn it off (most SCSI based solutions do), but in practice many ATA drives lie to you when you tell them to turn off certain features (like Write Cacheing) and leave them on.

Re:native command queueing

[garat]

Are there current SATA raid controlers even able to use Command Queuing? If not how long until it is possible. It may be a feature that couldn't be used at this stage but later revisions might.

According to storagereview.com the "upcoming Promise FastTrak TX4200" of a year ago uses it; as the article is over a year old I'd surmise there a plenty more.

Re:native command queueing

[garat]

Because that functionality is supposed to be handled by the RAID controller, and if the drive has it enabled you can end up losing performance as the two Command Queue Controllers fight each other.

From the excellent storagereview.com article: "TCQ must be supported by both the controller and the hard drive itself."

That's with TCQ being NCQ but for ATA-4.

Re:native command queueing

[hackstraw]

CPU load is also slightly higher when Command Queuing technologies are used.

I'm not sure if the command queuing is done on the physical drive (seems reasonable to me) or in the SATA driver (could be). But either way, the command queueing is not going to be a load on the CPU. The observed load on the CPU is probably because it is free from waiting on the disk and is going about its business doing what its supposed to do.

Re:native command queueing

[sumdumass]

and the FastTrak TX4200 is a software controled card. Sure it bills itself as a hardware controler but it relies on a drive in the host operating system to function. A real hardware controler will give access without a device driver installed. This is a real difference here.

Software raid controlers that require drivers to function can implement thru software the controls neccesary to use this. The problem is that not many enterprise level customers will have a software raid controler in thier systems. Hardware controlers can be update with firmware updates but i'm not sure if it is as easy as writing new software.

I know i didn't specificaly ask about software or hardware controlers. I guess i was asking from a prejudice about hardware controlers. The FastTrak TX4200 might be able to do it. I doubt how well it would work though. It might be interesting to see how it performs.

Sal Cangeloso is a moron

[laing]

The manufacturer specifically says to only use these in a RAID-1 configuration (mirroring). They have a reason for this: The error recovery mechanisim is abbreviated. So what does Sal do... He connects two drives in a RAID-0 configuration. Now his data reliability has gone to about 1/4 of a regular drive.

Re:Sal Cangeloso is a moron

[Jeff DeMaagd]

Yep, RAID-0 is the configuration you want to use if you want to risk losing data. You want RAID-1. I'm not saying that it's an excuse to have a proper back-up though, RAID merely keeps you going should a drive die.

Re:Sal Cangeloso is a moron

[fnj]

Damn right. For those who don't get it, RAID-0 is not even really RAID. RAID is Redundant Array of Independent Disks. RAID-0 does not have any redundancy! It should be NRAID-0 (Non Redundant Array of Independent Disks). Only an idiot would even consider RAID-0 for any purpose.

Re:Sal Cangeloso is a moron

[TheOrquithVagrant]

If you need speed, but don't care about data loss raid-0 is entirely appropriate. For example, I keep my games installed on a 4-stripe raid-0 volume, since it helps considerably with level load times, and should a drive fail, it's easy enough to reinstall the games.

Re:Sal Cangeloso is a moron

[kasperd]

I keep my games installed on a 4-stripe raid-0 volume, since it helps considerably with level load times

You could probably match the read performance or at least get close if you were using raid-5 rather than raid-0, and it would give you much more data security. It is on write performance you see the significant difference between raid-0 and raid-5.

Re:Sal Cangeloso is a moron

[TheOrquithVagrant]

What part of "don't care about data loss" did you fail to understand?
Why would I want to waste %25 of my volume's storage capacity to get better data security on something where I don't _care_ about data security? And no - raid-5 doesn't match raid-0 for speed even on reads, at least not in my linux software raid setup. No "probably" about it - I have a raid5 volume running on the same hd's, where I keep data I actually care about.

Re:Sal Cangeloso is a moron

[eatjello]

But how costly is a RAID-5 solution on a Windows box? You're not going to be playing Half-Life 2 on your Linux box any time soon, so you're going to need a hardware RAID-5 controller. When it comes to games, or Windows installs in general, I also run my setup off a striped array (2 disk). If something goes down, I can restore the game installs and OS from my file server, which _is_ RAID-5. It takes about an hour, unattended, to restore the images over gigabit LAN... one hour every 3-4 months (while I sleep) is well worth saving a few hundred dollars buying a PCI RAID controller card for my Windows box, which would probably saturate the PCI bus before I saw max gains anyhow (if I were to run RAID-5 from it, with 4 or more disks).

Re:Sal Cangeloso is a moron

[eparusel]

If your setup supports it, wouldn't an 8-disk raid5 be most efficient for your uses?

I don't know, a dedicated 4 disk raid0 for your games? Seriously? Is the load performance of your games that important?

*shrug*

Re:Sal Cangeloso is a moron

[GungaDan]

Not to niggle, but your assessment is incorrect. RAID-0 is equally as redundant as RAID-1. There are two disks performing a job that either one of them could do alone. The second disk is redundant. RAID-0 uses the redundancy for performance instead of continuous backup.

OK, so it was just to niggle.

Re:Sal Cangeloso is a moron

[Anonymous Coward]

There is nothing redundant about RAID-0. Do you understand what redundant means? Go back to MCSE school.

Re:Sal Cangeloso is a moron

[dbIII]

Only an idiot would even consider RAID-0 for any purpose.

If the consequences of total drive or card failure are very small it isn't a problem. I have a production machine with a couple of striped 200GB disks, but can afford to lose it for up to a week if necessary. With that system a full bare metal restore would only take a few hours unless I have to spend time finding a new card.

Re:Sal Cangeloso is a moron

[sld126]

RAID-0 is equally as redundant as RAID-1.

I think you're drunk...on niggles.

Re:Sal Cangeloso is a moron

[swillden]

If your setup supports it, wouldn't an 8-disk raid5 be most efficient for your uses?

Have you ever tried to cram eight disks into a typical PC? Four is tough. For eight you'd have to upgrade your PSU and if you're not using SCSI you'd also have to add extra controllers. Sure, it decreases the space "lost" to the redundancy, but unless you need vast amounts of storage, it's much cheaper to buy, say, four larger drives rather than eight small ones.

Nope, I think the most efficient use would be to have four drives and build multiple RAID arrays across them, using different RAID levels for different purposes. For example, in my file server I have four 200GB drives. By partitioning the drives and creating various RAID arrays across the partitions, then using LVM to meld them together, I have three volume groups, one that comprises 480GB of RAID-5 storage, one with 20GB of RAID-1 storage and one with 80GB of RAID-0 storage. From those volume groups I carve out logical volumes for my data. I can pick the performance and integrity levels appropriate for a given set of data.

Re:Sal Cangeloso is a moron

[ars]

Um, that's not true.

Raid-0 uses the redundancy for performance? Do you even know what raid-0 is?

Yes, there is a performance increase, but not even the slightest bit of redundancy.

Oh, and raid-1 has BETTER performance then raid-0.

Re:Sal Cangeloso is a moron

[kasperd]

And no - raid-5 doesn't match raid-0 for speed even on reads, at least not in my linux software raid setup.

What measurements do you have to back up that claim? Unless you are careful about it, you are going to end up doing both reads and writes, and it doesn't take many writes to reduce the performance of a raid5. I know sequential reads can be slightly faster on raid0 than raid5, but only as long as the disks are the bottleneck. If the PCI bus becomes the bottleneck, it will not matter anymore. And three disks are more than enough to saturate a 33MHz 32bit bus. For random access you shouldn't expect much of a difference between raid0 and raid5. Each logical sector exist only in one physical sector, so you can expect the same number of seeks on the same drives. And if you can sometimes save a seek because the head was already on the right chunk, it will be so on raid0 as well as raid5. In a few rare occasions raid5 may even be able to chose between reading the data and the parity because all the other sectors are cached, in this case performance could be improved by sending the read to the least busy disk or the one that would need the shortest seek. On raid1 random access reads can be even faster as the raid can always chose between multiple disks. If performance is the main priority I'd test a raid 1+0 with those four disks.

Re:Sal Cangeloso is a moron

[Anonymous Coward]

SW RAID5 is as fast as SW RAID0 as long as you have the same number of data disks. A 4 drive RAID0 is the same as a 5 DRIVE RAID5 set for read performance.

In your case if you have 4 drives with a RAID0 volume and a RAID5 volume on it the RAID0 set will perform better in reads.

Re:Sal Cangeloso is a moron

[m50d]

A raid5 array should read at exactly the same speed as a raid0 with one fewer drive. If it isn't there's something wrong with the implementation.

Re:Hey, I thought it was funny.

[Anonymous Coward]

Personal mod up to 2 yaro points!

Roland, anyone?

[Anonymous Coward]

Well, it seems the Roland experiment worked very well, so they've decided to move up to the Big Money.

NCQ..

[groovy.ambuj]

NCQ allows hard drive to reorder various commands/accesses to suit its current head position. Depending on your app you might not see a lot benfits from it e.g when you do serial access all the time but lack of it will certainly cause degradations when multiple apps are active. Also by using one big hard-drive instead of multiple smaller ones its putting all eggs in one basket. Mechanical problems are more frequent than magnetic ones for a hard drive..

looking for an inexpensive raid5 tower

[v1]

These buggers are hard to find for anywhere near decent cash. I've found one model that is fairly popular, going by several different names and brands, but nobody seems to have them in stock. They look like a GREAT deal and loaded with most or alll of the best features of raid5. (hot swap, live rebuild, live GROW, etc) Has anyone seen one IN STOCK anywhere?

Same exact models:

http://www.raidweb.com/fb605fw.html
http://www.micronet.com/General/prodList.asp?CatID =45&Cat=Product
http://www.firewiremax.com/fire-wire-1394-ilink/mi harasyfor5.html
http://www.pcrush.com/prodspec.asp?ln=1&itemno=779 19&refid=1057
http://www.cooldrives.com/firewire-raid-5-enclosur e-mini.html
http://www.topmicrousa.com/combo-205.html

same internals, different enclosure:

http://fwdepot.com/thestore/product_info.php/produ cts_id/657
http://www.cooldrives.com/fii13toatade.html

Everyone I call says they have them in stock. Then I ask them to check and they suddenly change their mind and say no it's not really in stock, (despite what their web page says) and they expect it in the generic "1-2 weeks". (retail-speak for "we don't know when it'll be in, please call back later")

Two of them actually told me they have yet to receive any of these units, so I don't think they've shipped from the manufacturer yet? (vaporware?)

Re:looking for an inexpensive raid5 tower

[John Napkintosh]

Seems like you could save quite a bit of money by going with something like this (assuming it was SATA you're looking for):

http://www.macgurus.com/productpages/sata/satakits .php

They have 2-, 3-, 4-, and 8-bay kits to suite your need. Get 'em with or without drives, cables, etc. The only drawback I see is the lack of a controller card (might have to go with something like the Sonnet further down the page). Then again, this may not be such of a drawback, since you're not stuck with a built-in RAID controller, in the event of it being a pile of junk.

$451 for an empty 8-bay with no cabling ain't too bad.

Re:looking for an inexpensive raid5 tower

[v1]

I've already got a pair of 8 bay towers, same design as your link but beige instead of black. Been using them as software mirrors, which has worked in the past but is becoming very cumbersome. I just was hit with a DOUBLE drive failure that very nearly cost me 250gb of data, so I am looking for a RAID5 self-contained solution. I could really use the improved efficiency of usable space in raid5 - 80% on a 5 drive system, as opposed to 50% on my mirrors. I also want a box that I stuff drives into, and plug one firewire cable into and it looks like a giant HD, with raid5 protection. This will allow me to move the array to another machine when needed. Can't hook it to my laptop if it requires a controller card!

Re:looking for an inexpensive raid5 tower

[Anonymous Coward]

Again, RAID is not an excuse to not back up your server. If you don't back up your server, you will lose data, no matter how many drives you have in a raid-5 array!

Re:looking for an inexpensive raid5 tower

[minion]

Seems kind of expensive for what they are. We use these at the office: http://www.infortrend.com/ which can be purchased from Adjile Systems.

Re:looking for an inexpensive raid5 tower

[v1]

Unfortunately this is a charitable organization. In other words, money is unfortunately a factor. Looking for the best protection for the buck. Right now a RAID is looking like the best way to go, as there is a bit too much data to cost-effectively back up. The current upgrade is a result of a DOUBLE failure on a mirror, which very nearly tanked 250gb of data. (you don't EVEN want to know what I had to go through to get the drive that still spun to mount back up...)

If we can get the cabinet cheap enough, we will be able to buy a fresh set of 5 drives for it instead of recycling our current drives, and can break the mirrors and use their drives for offiste backup. But that's a big "if" right now.

In a perfect world, money would not be a factor, but it's not, so it is.

Re:looking for an inexpensive raid5 tower

[v1]

Those use quite a few more drives than I require, but worth looking into. Shame they have no prices on their web site, will have to give them a ring on Monday and see what they can do.

In my brief browse I didn't see a big writeup on what those enclosures can do, though I did find several good descriptions of what those 5-bay units I found can do. They seem to do everything we need... firewire 800, no controller card, hot swap, hot rebuild, and though I wouldn't trust it... hot grow. That feature set has been going for $3500-ish a year ago, so it's really come down in price lately, which is why we're in the market.

Re:looking for an inexpensive raid5 tower

[pe1chl]

Recently I have been looking into adding about a TB of storage to my home network, and the different options:

- an external box like you describe
- an external box but with a small motherboard in it, and setting it up as a NAS system on the LAN
- the 5-drive cage that is in there added to my system
- adding bare drives to my system

For now I have just bought some large drives and added them internal to my system in an arrangement similar to what those cages use (drives stacked vertically into 3 5 inch bays).

I closely looked at the ARC-5010 cage that is used in those external cages. I think it is not a very recent design, and will probably be replaced by a newer type sometime, hopefully soon. That may also explain the stock problems.
This unit uses parallel drives only, although it has both a parallel and serial interface to the host system. The boxes you are looking at convert this to USB and Firewire using yet another bridge board.
I expect a new drive cage type would have:
- serial ATA for the drives
- serial ATA, USB and Firewire for the host
Making swapbays for serial ATA drives is a lot easier.
Also, the current model has performance bottlenecks that are below those of current drives. This may not bother you when you are considering USB and Firewire, but 5 drives directly attached to the computer will have an aggregate throughput that is way above what this unit can do.

In my case, I was not looking at hotswap. I don't mind when I need to shutdown the system to replace a drive, it is for home use (videorecording) only.
I wanted it to run on Linux, and that further limited the possibilities.
The units you are referring to would work on Linux, although I have read some mixed reports about connecting drives via USB (hangs etc).
Using a multi-SATA controller with an external SATA tower would work when the controller is supported. Several of the all-in-one solutions I encountered came with a controller that only works in Windows.
However, with external boxes with many drives there is a lot of cable clutter, and a new widget has been introduced: a SATA port multiplier. This allows about 5 drives to be connected via a single external SATA cable. It reduces the throughput, but with SATA 300 it should still be enough.
Unfortunately, Linux does not support those external port multipliers at all.
When you are using Windows, the least expensive solution could be to use an external cage with sata port multiplier and a single-port SATA controller in the system (whcih still can offer onboard RAID when you like that). However, that will not be a "storage in a box" solution. For that, using a NAS solution could be the better choice anyway, because you access it at a file rather than a filesystem level.
However, unless you assemble and configure it yourself (using a small PC motherboard and one of the available NAS solutions) it is going to be much more expensive. Ready-made NAS boxes tend to be quite expensive.

Re:looking for an inexpensive raid5 tower

[v1]

In my case it's on a macintosh (not windows nor linux) so sometimes I have as much problems with drivers as you do on linux. I'm looking for a nice self-contained unit that has just the one port to connect to, for maximum portability. High speed is not as important as data protection or portability. I don't think I am ready to assemble my own NAS box either, so the best bet for me right now looks like one of these mini raid5 boxes. One of the machines I'd like to cable up to is a laptop, so an external sata interface is out. I don't care much for USB even though I know 2.0HS is easily on equal footing with FW400, so I would prefer a firewire solution over USB. (use what you're comfortable with in this case)

Re:looking for an inexpensive raid5 tower

[pe1chl]

The ARC-5010 looks attractive in this case.
There was one other issue that I did not follow through: I wanted my discs to spindown when the data is not being accessed (I use my setup for videorecording so there are long hours it is not accessed). I am not sure if the ARC-5010 will allow the drives to spin down.

Anyway, in a RAID-5 setup the drives all need to be spinning when some data is accessed. This is a disadvantage over using pairs of drives in RAID-1, where only a single drive is required for reads and two drives during writes.
This may not matter much for "office" use. My system is in the livingroom and I prefer one or two spinning drives over 5. Although they are getting quiter all the time.

I am not familiar with the requirements a Mac would have on a NAS. The commercial NAS systems of course support the Mac, but they are a bit costly.

I hope you can get your ARC-5010 based unit delivered, as it seems an attractive solution for you.

Re:looking for an inexpensive raid5 tower

[Sketch]

Unfortunately this is a charitable organization. In other words, money is unfortunately a factor. Looking for the best protection for the buck. Right now a RAID is looking like the best way to go, as there is a bit too much data to cost-effectively back up.

--

I work for the Department of Redundancy Department.

Wow, so the Department of Redundancy Department can't even affford decent redundancy?

Re:looking for an inexpensive raid5 tower

[Anonymous Coward]

FireWire Depot does have the enclosures available and they normally ship within 24-48 hours of the order being placed:

http://fwdepot.com/thestore/product_info.php/produ cts_id/657

Network RAID?

[Eccles]

Is there a reasonable cost, relatively low power RAID-5 setup for home networks? I'd love to set up a file server with gigabit ethernet and RAID-5 to serve as the home directories for my multiple machines. Things like the Buffalo LinkStation are a step in the right direction, but no RAID, etc. Is my only solution a Celeron or Pentium-M based PC? If so, is it possible to set up such a system to act as home directories for a combo of Windows, Mac, and /or Linux machines?

Re:Network RAID?

[NekoIncardine]

... This would make an excellent Ask Slashdot. There's enough subtleties to make it a challenge to definitively answer, but it's general enough that other users might be interested.

Re:Network RAID?

[william_w_bush]

buffalo has a terabyte 4x250 drive raid capable of raid5 for 750g. has gige too, but a relatively slow processor, and though the box clearly states it supports nfs it doesn't and they don't plan to. if you're pure windows and just looking for a nice solution it sounds good though, very small and easy to manage. mac and linux will have to use samba, which is a decent bit slower, and without the same permissions, but how much slower depends on your workload. For things like a central backup for important data it sounds great. Costs about $1k, so it's not cheap nor ridiculously expensive.

Just a thought.

Re:Network RAID?

[canadiangoose]

I really think your best bet would be to use an old P3 or Athlom running software RAID on Linux or BSD. You can add several PATA/SATA cards to your machine and stack it full of drives and fans, and I think you'll find the performace to be acceptable. Of course, no software RAID can compete with an expensive SCSI RAID card with a dedicated XScale chip or whatever, but it's a heck of a lot cheaper.

It also depends what you want to be doing with it. I've played with both hardware and software RAID5 and home and at work. Software RAID offers excellend bandwidth, and seems to use very little CPU time. This is why I think a P3 should work. However, the seek time is terrible. Perhapse it has something to do with the RAID intelligence being located so much farther away from the drives than it would be with a dedicated RAID card. I've tried running an SQL server on soft IDE RAID on a dual Xeon 3.2, and it had the snot kicked out of it by a dual P3 700 with an ancient MegaRAID driven SCSI array.

As for running it as a home directory for Win/Mac/Linux, between Samba and NFS you should be just fine. You may even be able to go the fancy route and set up a few logical volumes as iSCSI targets and run your own SAN.

Re:Network RAID?

[Anonymous Coward]

I find people with stupid signatures about their low slashdot user id annoying and not worthy of reponses.

Re:Network RAID?

[xlsior]

One of the major reasons for the high price of most hardware RAID5 solutions, is the hot-swap backplane. If you are OK with a solution where you would have to shut down the server in order to replace a bad drive (which would be OK for most home use I would image), you can find some *very* cheap hardware RAID controllers ($50, for both ATA and SATA) that will do the job just fine...

Re:Network RAID?

[Anonymous Coward]

I use an old P3/600 with 4 250GB drives (2 raid 0+1 sets). File serving is generally not CPU intensive and I can achieve saturation on 2 100mbit eth lines while reading data from multiple machines. That is a relatively cheap setup but what do you consider cheap? The only thing I do at home that needs more bandwidth then my server may be able to handle is video capture and a second 200GB drive in the local machine I am doing my video on is more then adequate. My final video edit and any work I am doing gets copied automatically at night to the file server for backup which eliminates the expense and need for gigabit and and a faster raid setup.

A side note. In roughly two years with this setup, I've had three various raid failures or reduced raid functionality (drives marked bad that were not bad or a corrupt raid config). None of these problem were the drives themselves. The raid setup caused more problems then it has prevented so use caution when trying to achieve a "cheap" raid setup. Next time it happens, I'll probably get rid of the raid setup and go back to four invididual drives and using rsync between two similar sized drives via a cron job.

Assuming you really need Gb speed and want raid 5, your setup done reliably can get expensive quick. Don't forget to look into your data recovery options if you decide on hardware raid and your card fails!

Re:Network RAID?

[sumdumass]

Whats your idea of reasonable cost? You could probably do just that for less then $1500 on a PC and use linux with the varios SMB configs around. Or you might be able to use some windows version with cygwin exporting NFS or somethign. Any ways, $1500 sounds a little expensive for most (it is the price of a new midlevel gaming rig).

I'm sure there are some NAS stuff but they tend to be what i would consider pricy too. OF course reasonable cost is a reletive term so my idea might be lower then yours.

Re:Network RAID?

[Homology]

Is there a reasonable cost, relatively low power RAID-5 setup for home networks?

RAID-5 for home networks is a solution looking for a problem. RAID-5 is nice for minimizing down time, but for a home network that is very seldom the case.

You see, the problem is usually not that my harddisk failed, but that I need to get an older version of a file, or get a file I deleted by accident. RAID-5 is utterly useless for this. For most home users it's better to use something like rsnapshot and take daily/hourly snapshots of their main harddisk to other hardisks.

Re:Network RAID?

[Anonymous Coward]

It's been done a thousand times before.

Re:Network RAID?

[drinkypoo]

A Celeron isn't really amazingly low power. I'm thinking something more like VIA EPIA. You don't really need a lot of CPU as long as the only thing the system is doing is filesharing and handling the RAID. Generally speaking a fairly current CPU will knock the socks off of the RAID performance of even fairly expensive controllers because they really don't have all that much CPU on them. Using linux md or similar you can create whatever kind of RAID levels you want. Pentium-M would work fine, but the power consumption tends to be up there.

Just remember, if all you're doing is the filesharing and the RAID, you don't need a RAID controller. Unless you're on solar, the savings of not buying the controller might be enough to justify using a non-low-power processor ;P

Re:Network RAID?

[Samedi1971]

If you want low-power, buy a decent low-end setup and underclock the processor. I'm in the process of building a server (the 8-disk array is being formatted as I type). I started with an abit kv-81 with onboard video and gigabit ethernet. Then filled the two pci slots with 2 syba 4-port SATA adaptors. I'm using 8 250G western digital refurb drives I got from www.geeks.com for $80 each. For that price I'll risk the early failures and be prepared to replace drives. Before I found those drives on sale I was leaning toward the 320G drives in the article.

Slackware 10.2 is easy to get set up and running on a RAID-5 configuration (other than having to wait forever to format with a thorough read/write test). For under $1000 I've got a pretty decent 1.75TB media server, and that also includes 256M ECC registered memoery and a UPS. It doesn't include the case since I'm building a custom box for better soundproofing and cooling.

Re:Network RAID?

[aminorex]

It is perfectly feasible, however, to hot-swap SATA drives.

Re:Network RAID?

[aschlemm]

For a small SATA RAID setup for a small server I've been very happy running a system with a 4 port, SATA-150 LSI MegaRAID card. It supports RAID 0, 1, 5, and 10 in hardware. It's not the cheapest way to go but with some OEM SATA drives the card shipped with its own SATA cables so I was good to go. I've run both SuSE Linux 9.2 and 9.3 on my server and the LSI controller was supported out of the box. My needs were pretty simple and so I've been using two SATA drives in a RAID 1 (mirrored) setup. I guess I regret I didn't fork out a bit more money and get a 3rd SATA drive so I could do hardware RAID 5 instead and have more disk space available.

Dumb Drives

[Doc Ruby]

EIDE drives are the cheapest type. But AFAIK, each drive has a controller card onboard, which seems redundant when all the drives are being controlled in conjunction. Software RAIDs seem to have parity (pun intended ;) with HW raid controllers, but wouldn't a real "Made for RAID" drive have nearly no controller logic of its own (maybe just data separator and head/spindle speed/position calibration)? Lots of logic for controlling the RAID drive will be on the central controller card, or running on the CPU. So why have more on the drive? The cheaper the drives, the bigger the array at the same budget (shared overhead of common controller).

Am I correct, or are some RAID drive makers already doing this? Or have I just got all the controller:drive economics wrong?

Re:Dumb Drives

[Federico2]

Sorry but you are forgettin' one of the market laws that make EIDE disks so cheap: more you build, less they cost.
Hard drives built customized for RAID usage will never be cheaper than standard EIDE disks.

Re:Dumb Drives

[m50d]

The IDE chip on a drive is as simple as you can get while still being backwards-compatible and having a modicum of performance. If you made it any simpler the raid controller would have to know about the physical geometry of the drive. I'm pretty sure the circuitry is pennies of the drive cost.

Re:Dumb Drives

[Doc Ruby]

I'm talking about using the same EIDE drives, but factoring their controller cards into one in the RAID controller. That makes each drive cheaper. And the multiple drives mean more drives, which increases their economy's scale. So using EIDE HW for growing demand for RAIDs seems like a good way to grow the market for drives, better in fact than R&D for higher density individual drives.

Alpha Hardware

[lullabud]

I got an alpha hardware version of one of these bad boys and it came with a very handy extra that they assumed the end users wouldn't need: a de-bug port. Now I don't even NEED raid, but it's nice to have the option.

How about RAID on a hard drive itself

[Anonymous Coward]

Specificially RAID 0, mirrored. It would be nice to be able to split one of those oversized drives into a mirrored drive, using the opposite sides of the disk platter as mirrors. You'd get better reliability with the slight trade off of all that surplus disk space you never use.

Re:How about RAID on a hard drive itself

[LuckyStarr]

And what if the electronics of the drive fails? Or the motor? Or the drive-head actuator?

Sorry sir, but this is a bad idea.

Re:How about RAID on a hard drive itself

[dtfinch]

Not sure why this was modded interesting instead of funny.

Re:How about RAID on a hard drive itself

[Pop69]

Did you miss the Redundant bit in Redundant Array of Inexpensive Disks ?

The idea is that one drive going bad doesn't take out the whole array in mirror, not that you have it all on one drive you fool.

Re:How about RAID on a hard drive itself

[m50d]

And what if the electronics of the drive fails? Or the motor? Or the drive-head actuator?

You're in the same position if the controller fails in most hardware raid setups. As long as the platters are replaceable for when the first one fails, I think RAID1 within a two-platter drive has possibilities. Of course, the question then is whether you can make a drive with two user-replaceable platters easier than making two separate drives with half the capacity.

Re:How about RAID on a hard drive itself

[rholliday]

If the controller fails in a serious hardware RAID setup, you can replace the controller and reimport the configuration. Granted, you did say "most" hardware RAID setups, but I do not think that this is a good justification for a bad idea. Even if you can't reimport the config, sometimes you can recreate the array around the data. And if it's a mirror, just re-set primary and secondary.

Just because some hardware RAID configurations have the controller as a weak point doesn't mean that "RAID on a hard drive itself" makes any sense.

Nor RAID 0, for that matter. But that's another argument for another day. :)

Re:How about RAID on a hard drive itself

[m50d]

If the controller fails in a serious hardware RAID setup, you can replace the controller and reimport the configuration

True, assuming of course you can find one. But it would be the same with this removable platter drive - just take both platters out and put them in a new drive

Even if you can't reimport the config, sometimes you can recreate the array around the data.

I wouldn't want to be relying on this.

Just because some hardware RAID configurations have the controller as a weak point doesn't mean that "RAID on a hard drive itself" makes any sense.

It shows that having a solid state component as a single point of failure is an acceptable situation in some setups where the failure of the mechanical drives is still unacceptably high. A "raid on the drive" disk wouldn't be good enough for many setups, but I think there is a niche for it.

Nor RAID 0, for that matter. But that's another argument for another day. :)

The purpose of raid0 is to increase performance. Now that drive bandwidth has pretty much caught up with that of the bus it's pretty much obsolete, but it wasn't always so.

Go RAID! How real?

[RealisticCanadian]

Now, firstly, yay! I've loved the idea of RAID arrayed drives since I first heard of it about 2 years ago when my friend was writing linux drivers for a company's (obligatorily namelessness) RAID Card. The technology has some very high potential, but it has very much sat by the wayside. Seeing a major hardware manufacturer target a product to the RAID crowd means it's starting to get some industry clout, which I for one think is mucho overdue.

Secondly; however, the article seems very short on technical specs:
"In works better in RAID" according to.....?

Re: Go RAID! How real?

[MarkTina]

What do you mean sat on the wayside ? It's been out and about for donkeys years, I've been involved in storage for 9 years and it pre-dates me by a LOOOONG time.

Is it just me or....

[rongage]

Is it just me, or did this review stink for lack of proper testing and comparison...

If I were comparing this product and it's performance, I certainly would not be benchmarking a SATA based RAID setup against a single Parallel ATA drive. Something in this arrangement just doesn't seem... well, logical.

If you were really going to try to impress me with it's performance, then you would have to show me how it compares to "non-RAID" optimized drives of near simular characteristics. Show me how this drive performs against, say, Hitachi SATA 320 gig drives using an identical test rig. Also show me how this drive compares to 320 gig SCSI drives. Show me the results as JBOD, RAID-0, RAID-1 and RAID-5. You know, like the real world.

While the graphs are pretty, I'm afraid that this "review" it fairly content-free.

Re:Is it just me or....

[John Napkintosh]

It seems like the goal of their comparisons were to highlight the bang-for-the-buck value - the amount of 'high quality' storage you can get for the price - so it sorta makes sense that they'd compare it to an older technology that is a bit cheaper. For a company that's looking to get the greatest amount of space for the cheapest amount of money, 100/133 PATA almost certainly would be a consideration.

But it does get kinda confusing when they start throwing performance figures out there. Is it because performance was the goal? Or is it because the RAID level that doesn't sacrifice any overall drive space just happens to be considered a performance array level?

Re:Is it just me or....

[KillShill]

but it's fairly good at getting ad impressions, which is all it's designed to do.

most "reviews" on the web are are extremely basic done by people with little knowledge in the methodology of testing hardware/software.

it's useful in that it exemplifies how not to review products.

Re:Is it just me or....

[dtfinch]

SATA doesn't seem more expensive. For SATA and PATA versions of the same drive model, the price is usually always the same, or PATA is a bit more. It can actually cost more to produce PATA drives. If PATA is ever cheaper, it's only because retailers want to get rid of them. SATA costs about $0.45/gb for 250gb drives, a little less if you don't care about the brand name, or a little more if you want a RAID controller to go with it.

Hard Drives for RAID use

[Anonymous Coward]

Don't forget the "poor man's" RAID -> rsync. If you have 2 complete machines with 1 drive each, and you rsync them, you are *nearly* equal to one machine with dual power supplies, and mirrored disks. Of course, you probably would just mirror data, but you get the idea. Truly convenient if your backup fails one night.

Buffalo TeraStation

[chocolatetrumpet]

Buffalo TeraStation

Supports RAID 5.

I emailed if external USB hard drives could be added and swapped to a raid 5 array, and if it can be done "on the fly"...

but all I got was this lousy message:

"Please call (800) 456-9799 x. 2013 between 8:30 and 5:30 CT and our presales guys will be able to assist you."

I'm one of those weird people that would rather communicate in writing. Oh well - no sale.

Re:Buffalo TeraStation

[lordkuri]

Please call (800) 456-9799 x. 2013 between 8:30 and 5:30 CT and our presales guys will be able to assist you.

translation: call us on the phone so we can lie with impunity, and you can't prove it.

par for the course anymore

Re:Buffalo TeraStation

[nmos]

FWIW the answer is no. The external USB drives are seperate from the raid array. Also although you can read fat32 formatted external drives you cannot write to them in that or any other format a Win machine will know how to read.

What would REALLY make the drive RAID firendly

[jandrese]

IMHO, the biggest things manufacturers could do to make the drives more RAID friendly is to change the name (even with just a v1, v2, etc...) when they change platters.

Nothing is worse than buying a bunch of drives and a couple of spares and building the array and then discovering down the road that in fact one of your spares came from a different production run and has a slightly different (maybe 3 block smaller) geometry and can't be used on your array. Usually there is absolutely no indication on the box or the drive that one of your drives is different unless you decode the cryptic serial number.

For that matter, just printing the exact LBA count on the back of the box would be a huge boon.

This isn't limited to ATA drives either. I've seen it plenty of times in professional SCSI solutions too, especially as the arrays start to get older.

Re:What would REALLY make the drive RAID firendly

[swmccracken]

Recent promise RAID cards have a "gigabyte boundrary" mode, where they round the size of the array down to the nearest whole gigabyte.

This allows for minor variations in replacement disc sizes, at the cost of wasting some disc space. (It'd make a 250 gb array instead of a 250.23 GB one.)

Re:What would REALLY make the drive RAID firendly

[jandrese]

That would be fine except that sometimes a 250 GB drive is actually 249.7 GB. Drive manufacturers round up just as often as they round down. The best practice I've found is just to cut off the last 0.5% of the drive or so, but it's still annoying.

synchronized spindles?

[adrianmonk]

I would think if these drives are really designed for RAID (like other drives have been in the past), then they would have support for synchronized spindles.

The idea behind synchronized spindles is that in order to read data from a disk, you have to wait for the platter to come around part of a revolution for your data to become available, just like picking up your suitcase on the luggage carousel at the airport. How long you need to wait is a matter of luck, because the disk can be assumed to be in a random position when you decide you want your data. When you have RAID without synchronized spindles and you want data that's bigger than the stripe width (or when you're writing and need to update the parity), you have to wait for multiple disks, and they will tend to be spread out so that you tend to wait longer than if you were just waiting for one. With synchronized spindles, as soon as the whole group hits the right position, you've got what you're looking for, and you're done.

So, the point is, not having synchronized spindles tends to increase average access time, so having synchronized spindles is a desirable feature for a drive designed specifically for RAID.

Re:synchronized spindles?

[dtfinch]

On a mirrored RAID, having them out of sync can be better, at least in theory, if probably not in practice. There's no easy way for software to know where an SATA drive is in its rotation, but if you request the same block at the same time from both drives, one of them will respond first, and it will be sooner on average than if both drives were in sync.

Re:synchronized spindles?

[Jamesday]

If you're after data from two different stripe positions on a RAID 1 set, sending one drive to each place will get you the data faster than having both go to one place then both go to the other.

Re:synchronized spindles?

[Deflatamouse!]

Your maximum wait time is one revolution... whether you have one drive or if you have infinite drives, one revolution is the max time.

Having synchronized spindles seems a lot of work for a max saving of 1/2 revolution... in a 7200rpm drive, 1 revolution is 8.3ms, so 1/2 revolution is 4.15 ms... trying to synchronize spindles to save a maximum of 4.15ms seems to be a lot of effort put in the wrong place. You're better off going to higher rpms to squeeze out those few extra milliseconds...

Re:synchronized spindles?

[Zemplar]

" Your maximum wait time is one revolution... whether you have one drive or if you have infinite drives, one revolution is the max time."

Doesn't this depend on your RAID level? I would agree with you if you were discussing only RAID 1, but if you go to RAID 5 or RAID 0 are you not waiting for up to 1 revolution on each drive? Therefore, you could be waiting up to three revolutions on a three-disk RAID 5 or RAID 0.

Choice of RAID level should always consider data availability as well as drive latency and throughput.

Re:synchronized spindles?

[adrianmonk]

Having synchronized spindles seems a lot of work for a max saving of 1/2 revolution... in a 7200rpm drive, 1 revolution is 8.3ms, so 1/2 revolution is 4.15 ms... trying to synchronize spindles to save a maximum of 4.15ms seems to be a lot of effort put in the wrong place. You're better off going to higher rpms to squeeze out those few extra milliseconds...

First of all, it seems to me that synchronizing motors is fairly easy. The motors are presumably all locked to a clock anyway, so all you've got to do is run a cable so that one drive can slave its clock to a master clock, plus add in the ability to make adjustments until they're in the phase you want, and presto, you have synchronized spindles.

Second, why does making synchronized spindles have to be mutually exclusive with increasing the rotational speed? Sure, you can increase to 10,000 RPM or even 15,000 RPM, but drives don't tend to go much faster than 15,000 RPM, and if you want a boost beyond that, synchronized spindles can give it to you.

And third, the maximum saving is one revolution. There will be cases when you seek to the right track, and the data you want is right there with no waiting on one disk, but one of the other disks has just passed the desired point a tiny bit too early and you've got to wait for it to make basically an entire revolution before it comes around again. Yes, this case is uncommon, but it is possible. So, the maximum you can save is one revolution, although the average is lower. (Close to the 1/2 you mentioned, although it depends on usage patterns.)

Re:synchronized spindles?

[adrianmonk]

On a mirrored RAID, having them out of sync can be better, at least in theory, if probably not in practice. There's no easy way for software to know where an SATA drive is in its rotation, but if you request the same block at the same time from both drives, one of them will respond first, and it will be sooner on average than if both drives were in sync.

It's possible to make hard drives with synchronized spindles where the spindles are in sync but not in phase. You could put two hard drives precisely 180 degrees out of phase on a two-disk mirror (or 120 degrees out of phase on a three-disk mirror, and yes people really do have three-disk mirrors sometimes), and you'd get better performance than having the drives' relative position be random.

Re:synchronized spindles?

[Deflatamouse!]

You would be waiting for one revolution on each drive, you are right. However, this wait is done in parallel. That is the beauty of RAID. If the drives were accessed serially, then yes, you would have to wait the N revolutions... however, a correctly implemented RAID array should have the drives be accessed in parallel. Of course, I am not taking into consideration of the drive interface. If all the interface are through the same IDE card or if the PCI bus is busy, you would be waiting a lot longer.

Re:synchronized spindles?

[Zemplar]

Thanks for clearing that up. After I made that post I was wondering about just how "smart" most RAID controllers are and if those operations would be in parallel.

Knowing that you can put up to 32 SCSI drives on a single channel I would wonder about just how parallel most controllers can be. Granted most don't use that many drives and generally four or more drives is enough reason for a two-channel RAID controller for best performance.

,Picking nits I suppose when we are trying to cut one more revolution of a 10,000 rpm or even 15,000 rpm drive!

Western Digital is synonymous for crap

[Rev.LoveJoy]

In my whole IT career (some ... christ ... 13 years now) I have seen no other vendor of HDD that comes close to WD for sheer volume of failed drives (Maxtor is a distant 2nd). That they resort to cheap marketing gimmicks like this (1 million hours mean time between failure, puhleeze, these are the people who pioneered the 1 year warranty) is only so much more indication of their propensity to manufacture garbage.

Buy their gear if you must but I would not put my data on it.

-- RLJ

Re:Western Digital is synonymous for crap

[rizzo320]

Wow, did Western Digital plot to have your family killed? What a vendetta!

All hard drive manufacturers have gone through cycles of poor quality and reliability. Maxtor, Seagate, IBM/Hitatchi (remember the "DeathStar") have all had the same problems. In all my years of repairing and building desktops, I can say I have had the most problems with Seagates and (the now owned by Maxtor) Quantum drives. If you ask someone else, they'll give you a different answer too.

This drive has a 5 year warranty. Most other Western Digital's have a 3 year warranty, even if you buy the OEMs (in most cases). And read the articles above for what 1 million hrs MTBF means!

Re:Western Digital is synonymous for crap

[dtfinch]

I personally trust WD more than I trust Maxtor, but all manufacturers have bad years and bad models. This year I only trust Seagate, on only certain specific models.

Re:Western Digital is synonymous for crap

[Anonymous Coward]

In my whole IT career (some ... christ ... 13 years now) I have seen no other vendor of HDD that comes close to WD for sheer volume of failed drives (Maxtor is a distant 2nd).

And, what do you make of the fact that every single person in IT tells exactly that same story, but each one mentions a different brand?

Could it be that, from the perspective of someone dealing with these units in bulk, modern hard drive technology just isn't as reliable as it should be? Could it be that there are so many things that can go wrong with this kind of mechanical device that every brand seems to have good years and bad years and good models and bad models (sometimes even entire bad plants)? Your belief that WD drives are somehow exceptionally worse than the other brands is a prime example of cargo cult science.

Re:Western Digital is synonymous for crap

[twistedcubic]

For the benefit of the community, can you say which models?

Re:Western Digital is synonymous for crap

[dtfinch]

I'm still pretty new to drive shopping, but I'm thinking of the Barracuda 7200.7 at 200gb or less, and 7200.8 at 250gb or less, according to statistics and reviews from storagereview.com (registration required).

Old news

[Anonymous Coward]

I've had two of these drives in my server at home for almost year now. This is news??

Re:SATA version may be new, but features are not n

[FLaSh SWT]

No kidding. I've had a pair of 250GB IDE Raid Edition drives for over a year now.

Well my guess would be

[Sycraft-fu]

That the RAID card handles it. Please remember I am just guesing here, I don't know. However I know there used to be PATA IDE RAID cards that did this. The discs didn't support any kind of special reading, they just processed requests in order. Ok, no problem, the controller, which had a processor, RAM, etc did all that. It would implement scatter-gather and so on. Basically it was a SCSI RAID controller with IDE connectors instead.

So perhaps the thought here is since you have a controller that'll handle it, leave the features off the disk. PErhaps it's cheaper, perhaps it works out to be faster, perhaps NCQ qould interfere with the RAID card's operation. You know, maybe the RAID card queues up operations and then dispatches them to the disk in the order they should be executed,b ut the disk then queues them up and does them in an order that makes more sense to it, but less to the actual RAID.

Or, maybe they are just ripping people off.

3 platters

[dtfinch]

I'm no expert, but I look forward to mostly buying 2 platter drives from now on. Early failures seem to double when you add a third platter, and 5 platters is just scary. You can get 250gb SATA 2 platter Seagate drives for about $110 each, which seem to have a great record for reliability so far. But when I need real SCSI reliability I'll just get a real SCSI. The warranty for most SATA drives may be 5 years, but usually it's void if you put it in a server.

Enterprise level ?

[MarkTina]

Hmmm ... NOPE! ATA is good for 2nd teir storage but I wouldn't dare put anything "enterprise level" on it, just far to slow at the moment.

Time for integrated PCI-X controllers

[quazee]

There have been speculations about all these 1207 pins in AMD's SocketM2 (due somewhere in 2006) will be used for an integrated PCIe controller.
http://www.theinquirer.net/?article=24756

Hah!

[Anonymous Coward]

Not if you are super-l33t and run RAID-0 striping (which is apparently what these people think "RAID" is). Then it's like using two bullets in Russian Roulette.

Re:SATA version may be new, but features are not n

[Mechcozmo]

I bought one to replace what I thought was a bad drive in a RAID configuration about a year ago.

"Yeah, boss... drive, uh, died. I'll get a nice new one-- let me... uh... take this one home. I think it's, y'know, dead. Honestly this time."

Re:full article mirror & comment

[LiTa03]

I'd say.. avoid like the plague

There even was a /. story about it.

Here's the direct link about the AAC driver mentioned in the story.

back to the testing lab and do it right!

[Anonymous Coward]

I really wish there was some sort of standardisation of reviews such that ones like this could be filtered out. How it got on slashdot is beyond me!

They review a RAID edition drive yet don't even test it in RAID5! Unless reviews are thorough how are we supposed to draw anything but the vaguest conclusions. This reviews testing set should have included all these combinations:

- software vs hardware RAID solution (including hybrid semi-hardware cards like RocketRaid 1820A)
- 2,4,8 drive tests for RAID0,1,1+0,5
- synthetic tests such as the one they used or HDTach or similar as well as real world tests such as a database benchmarch, file server test

i hate to rant but these thoughtless reviews really are a waste of time.

NCQ is generally a good thing

[pensivepuppy]

I think the slashdot poster got it wrong. NCQ is generally a good thing. Without NCQ, the only way to get really good random-write speed is to turn on write-back caching, which is unsafe. THere's no good way the raid controller can tell otherwise which block would be best to send to the drive next, based on where the head is - only the drive knows that, and NCQ lets the drive order the writes best.

From what I understand, when writing to ATA drives, you have three options:
    - writeback caching on, in which case the raid controller can't really tell whick blocks made it to the drive and which ones are still sitting in the drive's (volatile) RAM.
    - No caching - send a block, wait for the drive to write it, send another. Unless the raid controller can really figure out where the head is and time it just right, this will be slow, because the writes will have to wait for the block to move under the head each time.
    - NCQ (which is in SATA only). The controller gives the drive a lot of blocks to write, the drive writes them in whatever order it chooses, and sends back acknowledgements for each block after it is written. These acks let the controller mark that block in its cache as clean.

If you compare NCQ and write-back caching, you'll probably find NCQ is slower due to the overhead of acknowledgements. If you don't care about the safety of the writes, then write-back caching is probably the best choice. If you do care about the data integrity, then NCQ is probably the best choice.

Some vendors (apple xserve for example, which doesn't support SATA), say you should use write-back cache but use a UPS to make sure the drive's cache is not lost. THis is not a "mission-critical" solution. Doing this means that you've now introduced single-point (the power connection between the UPS and the drive), where if there is a failure, you can wind up with bad data. Since RAID is usually intended to prevent data loss due to single points of failure, this is not a serious solution.

Re:NCQ is generally a good thing

[drmerope]

No. No. NCQ is bad in a RAID setup. The reason is that NCQ means that the head is in unpredicable place. This means that on a RAID5, each write operation will complete in a time determined by the worst seek. Conversely, most RAID controllers can reorder their own write operations. Moreover, they usually have their own caches--and battery backed ones at that.

Therefore, NCQ tends to offer no performance advantages, needless cost, and can occasionally do screwy things.

I don't even think the mainline SATA raid controllers 3ware, LSI MegaRaid, support NCQ on their array...

Your single-point of failure analysis is bogus. RAID setups do have a single point of failure: the RAID card for instance. Its all about probabilities. Disk crashes happen to be likely, other failure modes much less so.

SATA CRASHES

[AyeRoxor!]

I have a serious question. I have had *VERY* bad experience with SATA drives. I had a two-drive SATA raid0 for temp and cache and video processing. One drive crashed within 2 years; click of death. So of course the raid is gone, but I reformatted the remaining drive and used it by itself. Gone in under another year.

I dont remember the manufacturer but I realize it could just be a bad manufacturer.

BUT just recently the SATA drive in a friend's factory-spec dell crashed hard. No click-death but seems like serious controller damage/failure; will boot into safe mode with command prompt in about 20 minutes. Won't DIR, MD, etc. I threw int in another machine on the second SATA channel and results were the same so it was the drive.

This is the third SATA drive I've seen crash hard in the last 2 or so years.

Am I just having bad luck?

Re:SATA CRASHES

[lachlan76]

SATA is just different controller hardware. The drives are the same.

Not slower than SCSI

[Jamesday]

Since I wanted some facts, Wikipedia ordered two systems for database service, both dual Opterons with 4GB of RAM and six drives. One with 10,000 RPM SCSI drives and one with 10,000 RMP SATA drives. The SATA system, without NCQ, was generally faster and ended up with a higher proportion of the site load assigned to it. The SCSI system was sometimes faster in mixtures which included lots of writes with lots of reads and that made it lag a bit less in replication of bulk update operations, so newer systems have been SCSI. If more drive bays had been available, adding another couple of SATA drives would probably have made the SATA set faster for that case as well and still cheaper.

If lower access times are needed, SCSI drives beat SATA drives just because you can only get 15,000 RPM with a SCSI interface. May also make sense to have 15,000 RPM drives if you're already spending a lot of money on 16GB of RAM.

The question about this drive which interests me is whether drive write caching can be easily turned off and will stay off, so you don't lose database data when the database thinks the data has been flushed to the surface but it hasn't really been flushed. If you can't do that, it's unsuitable for a lot of database work - certainly unsuitable for use with RADI controllers with battery backed up write caches, where you have the battery to make sure you don't lose cached data if the power goes off. Anyone who things colo power and UPS will protect against loss of power hasn't suffered enough yet... :)

sounds more like READ

[Anonymous Coward]

Redundantly Expensive Array of Disks.

RAID 0 is fine, in its place

[Jamesday]

So, I have seven database servers, all with identical copies of the data. Do I really care if I lose all the data on one of them because one drive in a RAID 0 set fails? The completely redundant systems do the job better than any RAID setup can.

You consider RAID 0 when you don't care about losing the data if there's a drive failure and want the benefits of striping and the extra space available for a given number of drive bays, compared to other RAID levels. RAID 5 can get you some of the space but it's slower for database work.

easy external (or internal) raid5

[Anonymous Coward]

The units you listed (at least the first few) look like a drive cage inside an external enclosure. I do not know where you can buy one of those pre-assembled, but you can get the three major parts easily enough and just put it together yourself.

5 drive cage (built in raid - IDE/SATA with IDE drives):
http://areca.us/products/html/ide-ide.htm

1394 to IDE bridge board:
http://www.granitedigital.com/catalog/pg19_firewir ebridgeboards.htm

external 4 drive enclosure:
http://www.macgurus.com/productpages/scsi/mgscsien closures.php

if you don't already have drives, you could look here:
http://www.newegg.com/ProductSort/Category.asp?Cat egory=15

Alternately, you could get the areca unit and mount it inside the computer - just check that you have 3 free 5.25" drive bays without any protrusions.

I recently picked up one of the SCSI to SATA areca raid units and it has worked well so far - one of the fans went south, and they are shipping out a new one.

RAID Levels

[DragonHawk]

RAID-0 is equally as redundant as RAID-1. There are two disks performing a job that either one of them could do alone. The second disk is redundant. RAID-0 uses the redundancy for performance instead of continuous backup.

Who the heck mod'ed that up?

Say you have two identical disks striped without mirroring ("RAID 0"). The data is interleaved across the two disks. You usually get higher performance. You always get twice the disk space. There is no redundency here; your disk space is increased. Your reliability is much lower. If you loose either disk, you loose your entire logical volume.

Say you have two identical disks, mirrored ("RAID 1"). The same data is written to both disks. Data storage is redundent but reliability is increased. You can loose either disk but still have your logical volume. You usually get higher read performance (reads can be interleaved).

With mirroring, writes have to be done to both disks. Performance is complicated. Depending on the implementation, writes can be faster, slower, or the same as other RAID levels. With shared I/O paths, writes are slower then a single disk (you have to write one disk, then the other). In theory, if you have parallel I/O paths, writes should be the same as a single disk, but slower then a striped array (like RAID 5). That's because with striping, you can distribute the writes across multiple disks. In practice, many if not most RAID implementations (even those with dedicated RAID processing hardware) lack the processing power to make this happen. So mirroring is actually faster on writes then RAID 5, because no RAID calculations need to be done.

RAID 3 is striping with a dedicated parity disk.

RAID 5 is striped partiy (both data and partiy are distributed over all disks).

The best performance and redundency typically comes from using a striped array of mirrored disk pairs. You can loose either disk in a pair without loosing data. You distribute reads and writes over all disks. No fancy XOR calculations; just simple mirroring and interleaving. Very expensive, though.

Re:CAR CRASHES

[GigsVT]

I have a serious question. I have had *VERY* bad experience with cars. I had a two-car setup for running to the store and doing errands and picking things up.

One car crashed within 2 years; tires went flat and I couldn't steer. So of course the car is gone, but I got in the remaining car and used it by itself. Gone in under another year.

I dont remember the manufacturer but I realize it could just be a bad manufacturer.

BUT just recently the car in a friend's garage crashed hard. No flat tires but seems like serious damage/failure; the engine will start with about 20 minutes of cranking. Won't drive or turn, etc. I threw int in another garage and results were the same so it was the car.

This is the third car I've seen crash hard in the last 2 or so years.

Am I just having bad luck?

FYI

[Anonymous Coward]

It's "etc.", not "ect.".

Still leaves one big question....

[Casandro]

_Why_ are they "made for RAID use"?

I mean why is it that they are claimed to be made for RAIDs? The article claims some speciall error correction without explaining it. OK, you might think that the target audience might be to advanced for that.
On the other side he reviews hardware under only one OS where much of the performance is actually limited by the drivers. (There are many popular commercial OSes where gigabit Ethernet cards run at only a fraction of the speed of other OSes) IMHO that's a very unprofessional way to evaluate something like that.

Re:CAR CRASHES

[AyeRoxor!]

You're a dumbass.

If your car is crashing itself, while you're away, 3 times with 3 different cars, I'd see what the common variables are.

Let your daddy back on the computer, honey.

RAID Explained

[FauxReal]

From RAID Explained. (See link for info on other RAID configurations.)

RAID 0 uses a method of writing to the disks called striping. Let's assume you have a server with three drives of 500 MB, 1 GB and 2 GB. Normally a server would treat each of these drives individually. By incorporating striping, the system would see all of the drives as only one drive for a total of 3.5 GB. Big deal, you say. Wait, there's more.

When the system writes data to the disk, the RAID 0 striping kicks in and automatically distributes the data across all three drives. Part of a file (chunks of data) will be written to the first drive, the next part to the second drive, the next part to the third drive and then it starts all over again until the entire contents of the file have been written.

What this does is greatly increase the speed of the reading/writing process. If you have two drives on your server, it increases the speed by about 25%. If you have three drives, it increases the speed about 33%. When you consider that the main task a server is performing is reading and writing data, any increase in speed is highly welcome.

Besides increasing speed, the other benefit is that the drives can be of different sizes.

Because RAID 0 only writes the data once, it does not achieve data redundancy. If one of the drives fails, the entire system has to be restored because all files are split or striped across all drives.

Because there is no data redundancy, there is no loss of disk space.