Slashdot Mirror
Going Deep Inside Xserve Apple Drive Modules
adamengst writes "If you've had an Xserve drive fail, you may have considered saving some money by putting a replacement drive inside its Apple Drive Module. That may be a false economy, though. TidBITS explains why, while pinning Apple down on exactly what goes into Apple Drive Modules and why they cost so much more than bare retail drives."
Is this like how SCSI drives have special pixie dust on the platters that ATA drives don't, and that makes them more "enterprise-y"?
I want to delete my account but Slashdot doesn't allow it.
Slashdotted already!
I was going to complain that this is not a very interesting story for 98% of Slashdot, who has never seen an XServe and is happier for it, but since the link is already slashdotted, I guess I should complain about that instead.
Four comments in, and the server is pooched. Now, to keep this somewhat relevent, I had a hell of a time with hard drives when I tried to get Yellow Dog Linux running on an Apple Network Server, oh, eight years ago. Let alone the BIOS and stuff; oh, the hoops you had to jump through to get that to go!
Vintage computer games and RPG books available. Email me if you're interested.
While I hate to copy it, the server being pummeled and reporting errors for 9/10 requests doesn't lead to ad revenue either, so here goes:
About a year ago, we bought an Intel-based Xserve with a pair of 80 GB SATA drives to act as our primary Web server. When the boot drive went flaky on us in October 2008, we were able to recover from the backup on the second drive and off-site backups, if a little shakily (see "TidBITS Outage Causes Editors Outrage", 2008-10-07). But although we were able to bring the machine back online, we didn't trust the drive that had failed. Since the Xserve has three drive bays, the obvious solution was to purchase another drive. Sounds simple, doesn't it? Not so much.
You cannot buy a bare hard drive and insert it into an Xserve, as you can with a Mac Pro (and having just added a drive to my new Mac Pro, I can say that Apple did a stunningly nice job in making it easy to add drives, especially in comparison to the awful approach they used in the Power Mac G5). Instead, Xserves require Apple Drive Modules, which are custom carriers containing drives.
For users accustomed to buying inexpensive hard drives, Apple's pricing on the Apple Drive Modules comes as a bit of a shock. An 80 GB SATA ADM costs $200 from Apple, and a 1 TB SATA ADM costs $450. In comparison, a bare 80 GB SATA drive can be purchased for a measly $35, and a 1 TB drive is only about $100. That would seem to point toward buying a new SATA drive and swapping it into the bad drive's ADM. However, when I started down that path, a number of problems arose, such that I bailed on a quick solution and simply purchased a new 80 GB SATA ADM to replace the bad one.
First, I wasn't sure whether my Xserve had SATA drives, as I thought, because System Profiler on the Xserve shows nothing on the SATA bus, instead including all drives on the SAS bus. (SAS stands for Serial Attached SCSI, and is a high-performance data transfer technology that supports fast SCSI drives and is downward compatible with SATA drives.) After some discussion with knowledgeable folks on the MacEnterprise list and careful reading of the drive details in the SAS section of System Profiler, it became clear that both SAS and SATA drives are shown in the SAS section, with SATA drives having "ATA" as the Manufacturer, and showing "Yes" in the SATA Device line.
Second, once I knew that I had SATA drives in my ADMs, I started investigating if there were any gotchas involved in replacing the drives. There turned out to be surprisingly little hard information about this, with some people having replaced an ADM's drive with no trouble and others experiencing performance or reliability issues. I did find a few discussions about how replacing drives isn't recommended, but giving no solid sources.
Confused, I contacted Apple to discuss why ADMs are so expensive in comparison to bare drives, exactly what an ADM does, what Apple recommends users do with failing ADMs, and whether or not replacing a drive in one is a good idea. That conversation revealed a great deal of interesting information about the ADM and shed some light on what people with flaky ADM drives should do.
Drive Selection -- The most important fact to know about ADMs is that Apple doesn't use just any drives. We've all benefited from the amazingly low cost of storage. But whenever manufacturers compete on price, they cut corners every way they can to reduce costs. Although drive reliability is generally good, everyone who buys bare drives regularly has a drive vendor they refuse to patronize due to bad experiences in the past. (As is often the case, these people all hate different vendors, depending on which one was having a bad run at any given time.)
Since the Xserve is designed to be in constant use - 24 hours a day, 7 days a week, for years at a time - Apple doesn't use the least expensive drives available, since those drives are designed for more normal duty cycles in desktop computers - 8 to 10 hours per day, with variable use during that time. Instead, Apple wor
Apple dude discovers that servers use, well, server class HDs and they cost more than normal ones.
Oh, and the 'sleds' that hold the HDs have some LEDs (cool!) and a controller board to work with the cooling system.
Like pretty much every other half decent server then.
I want a list of atrocities done in your name - Recoil
What a long winded justification for why apple can charge over twice as much for some "enterprise level" equipment.
I'd rather just double up on equipment for the same money.
They are gonna need more drives then. I think this was part of Apple's plan all along.
Well, there's spam egg sausage and spam, that's not got much spam in it.
Sarcasm fail :(
Any life is made up of a single moment, the moment in which a man finds out, once and for all, who he is.
Wow, those Apple Disk Modules are cheap! A 1TB SATAII 7.2K RPM disk module for an EMC CX3 SAN runs about $1500. But I think they get to high grade the drive makers' inventory since they suggest only 1 hot spare per 30 disks.
The extra money is for the blessings of Jobs that make your hard drives last longer as they are and also run soooooo cool
The gist of the article is "We asked Apple why they're more expensive, and took their word for it." It's just regurgitated marketspeak about how Apple tweaks the firmware for the optimal performance, has special rubber on the grommets of the ADM that is specific to each drive to reduce vibrations, and how off-the-shelf drives are unreliable, slow, noisy, and hot.
They don't make an effort to verify this information at all. Because Xserves won't run with commodity drives, they can't do a proper comparison to determine how much is truth and how much is smoke-up-the-ass from Apple. This is such an astroturf article, it doesn't even pretend to be anything otherwise.
... nicely. Quoting TFA - "About a year ago, we bought an Intel-based Xserve with a pair of 80 GB SATA drives to act as our primary Web server. When the boot drive went flaky on us in October 2008, ... "
Welcome to pragmatism and reality - Drives fail all the times. So use cheaper drives in redundant mode, replace them with cheaper drives when they fail. You would have saved good amount of money even if the cheaper drives failed three time more than the costlier ones. (450$ for 1TB vs $100 for 1TB - from the same article.)
The summary makes it seems that there's no rhyme or reason why Apple charges more for their HDs and why can't the consumer simply replace it with a standard SATA HD. If you RTFA, it goes into a long list of reasons why. Whether you accept Apple's reasoning is another matter.
To begin, XServes use a HD module called ADM rather than simple HDs. On the new MacPros, they also use a module but those modules are designed to replace the HDs inside. For the XServe you apparently can't get a bare drive alone, you have to replace the whole module. The author begins to list the reasons:
And then the author concludes:
Well, there's spam egg sausage and spam, that's not got much spam in it.
If your OS could benefit from custom firmware...for example if you file system writes in certain sized blocks. I can see that being a case for specialized hard drives. But does that really account for the cost? If the drive's firmware is flashable, let the customer flash it to perform better with their OS choice.
The rubber grommet thing? Now that's some excellent bullshit. You are really telling me that someone spins up the drive, records the vibrating frequency, then selects the appropriate rubber grommets from the bin, then assembles the harddrive caddy? The bullshit flag is on the field, 10 yard penalty - roughing the truth. Again, even if that DID happen, does that justify the increased cost? I doubt it.
In the end, you still run a sluggish GUI on a server. fail. I bet if you ran your website on a stripped down *nix server, on a $1000 machine, your ass wouldn't be slashdotted right now.
THL phish sticks
Having worked in the disk mines of IBM many years ago, the SCSI disk controller is somewhat your pixie dust but the real reason is the disks, heads and other parts for the SCSI drives came from IBM's best manufacturing facilities. The deathstar ATA drive's parts came from the lesser manufacturing facilities. In theory a SCSI disk should not be much better than ATA but the reality is the best made, more reliable parts go to the high end more profitable products.
"Apple also told me that the rubber grommets that hold the drive to the ADM carrier are chosen specifically to match each drive's vibrational characteristics."
"Yes, sir. In order to reduce vibration we use the finest of synthetic compounds to minimize noise so that your cold room droogies won't have to suffer a higher level of acoustic trauma."
Get real.
Every major manufacturer re-labels drives for inventory and warranty purposes. They also use custom firmware to identify the drives for the same reasons. Special grommets? If you have worked with ADMs you know those grommets are extremely thin, shred to bits if you try to reuse them -more likely you lose them taking the thing apart, and are there more to keep the screws from coming loose than anything else. I have replaced drives in ADM modules before with RE drives -because the drives were mirrored- and haven't had a problem. If you are going to replace the drive in a server -a piece of mission-critical equipment- with the cheapest bulk OEM drive you can find, you will have problems.
Sig this!
Been happening for years.
Back in the late 80's, in addition to my dev job, I admin'ed a Motorola Delta 3600 box. We were looking for a little more space, manual said that it would take a Seagate ST-251N 40MB SCSI drive. So we bought on off-the-shelf.
It wouldn't work. It turns out that Motorola had custom firmware for those 251Ns.
So it's been going on for at least 20 years.
Fascism starts when the efficiency of the government becomes more important than the rights of the people.
The special ingredient in XServe disk drives is... love. :-/
HP and EMC does the same thing. HP will charge you $500 for a 1TB SATA today and we just paid EMC $800 or so per 500GB drive for a bunch of drives.
one nice benefit is the support. HP has a proactive failure warranty. if it flashes and alert that it thinks the drive will fail you call them and you have a new drive arrive by UPS the next day. EMC will come out within 4 hours to replace it.
and they are guaranteed to work with the brand name RAID controller that is the same brand as your server. you're paying for the testing and special drivers knowing that everything you buy will work together and you don't waste time calling support and playing musical telephone
Go to any server manufacturer's website (or a retailer if they sell through the retail channel). Dell, HP, IBM, I don't care. Any of them.
Price up equipment sold specifically for servers. Note particularly the price they charge for a larger/faster hard disk.
Go on, I'll wait.
Right, now go onto your favourite cheap & cheerful parts supplier and look at how much they charge for a hard disk.
Is it really the exact same disk with that much price discrepancy? Well, I (along with a lot of sysadmins) would dearly love to believe that it isn't. Whether or not that's true I honestly couldn't say.
What I can say is that if you do go out and buy the cheapest disks you can to populate the server, warranty support from the OEM is going to suddenly become "Oh, you plugged some random disk in? Go away and come back when all the disks are from us". Which starts to look rather expensive rather quickly when the RAID's knackered and you need to resurrect the system as quickly as possible. If your job is on the line, it's soon looking even more expensive, and nobody wants to say "I was sacked from my last job because I cut one too many corners on a system that was critical to the business" in an interview.
It's not so much of a problem for the Googles of this world who write their own applications to live on huge clusters which have component systems being added and removed all the time. Most of us, however, don't have that luxury.
Dell does the same thing with their poweredge drive modules and there are no electronics on the back, the sata connections go directly into the backplane . Prices are ridiculously out of whack. A 1TB drive is like $600 for a market priced $100 drive and a $10 piece of plastic and metal. This is why we always purchase the smallest 80 or 160GB drive module and put whatever SATA drive we need in. It's really stupid as are the idiots who purchase the larger storage modules.
This retarded fluff piece aside, the reason people buy (and pay a premium) for oem "blessed" hard drive replacements is because they JUST FUCKING WORK. If I save $100 on a hard drive, but spend two hours dicking with the raid controller to get it to play nice, or find out that it is in fact 2 mb smaller then the other drives, and now the raid won't rebuild, or has some firmware issue that I now need to rig up something to update, etc. I've lost money.
There is value in having everything already tested, and all your equipment in a "supported" configuration. When you have problems it makes it that much easier.
The fact that this article was apparently written by someone who does not know the difference between SAS and SATA makes it completely worthless. Clearly they are not qualified to admin the server they do have, much less write articles about the technical benefits of apple drives over other replacements.
My guess the custom firmware has more to do with Apple's monitoring and management tools for the Xserve. That being said, I have swapped out the ATA drives in a Xserve G4's drive modules with success. I went from the included 120GB drives to 250GB drives. I ran them for close to three years. I finally retired the server after purchasing a new Intel based Xserve to replace it. The reason the SATA drives are listed under SAS on the Intel Xserve is because they are connected to a SAS drive controller that is backwards compatible with SATA drives. I have two G5 xserves. I may try upgrading the SATA drives in the ADM's this summer since these are now out of warranty and are being moved to a less critical role.
I've never seen an Xserve, but it seems to me that if it is truly a file server, the drives themselves are probably hot swappable and that explains that increase in cost over a standard drive. The summary mentions something about "drive carrier". I read that and I picture the drives on my Proliant servers (both SCSI and SAS). They have special carriers and can be hot swapped while the server is still up (when running on the RAID module).
Guys drives dies in Apple server hardware.
Guy looks into buying retail drive for replacement.
Guy asks forum members for advice and decides to call Apple
Guy calls Apple wanting to know why their drives ar 4x+ the cost of retail
Apple gives Guy song and dance about magical marketing BS
Guy falls for BS and tells everyone else to follow magical marketing BS
Honestly what did he expect to learn by calling Apple? Call any manufacturer, tell them you want to use cheaper 3rd part parts instead of their overpriced parts and be prepared for a load of horse shit to flow from the phone. I worked in the IT department of my college for an elective credit way back when. The head IT guy almost never bought OEM stuff if their was a cheaper retail part that would do the job. He insisted its just a 3rd party part with company logo stuck on it sold at a 3-5x mark up. He would rather used the money he saved for better things like new equipment or upgrades. Never had any problems.
By propagating the '... fail' meme, you show how stupid you really are.
Go back to fark.
The article says that drives with mapped out blocks that don't work in a RAID work just fine in a Drobo. Actually, the Drobo is a RAID. It automatically configures itself to use either RAID 1 or RAID 5 depending on how many drives you put in it. These schemes have redundancy and therefore some robustness to them. Apple tends to use RAID 0 by default(which isn't really RAID and has no redundency) to improve performance and give the most possible space. When you do that, you really do need to be sure that you have good drives because the array will fail if either drive fails and you better have good backups.
I can hardly believe this article got posted on Slashdot. It's a bit disconcerting. However, I would like to reference one paragraph.
We've all benefited from the amazingly low cost of storage. But whenever manufacturers compete on price, they cut corners every way they can to reduce costs. Although drive reliability is generally good, everyone who buys bare drives regularly has a drive vendor they refuse to patronize due to bad experiences in the past. (As is often the case, these people all hate different vendors, depending on which one was having a bad run at any given time.)
The bottom line is, you buy a cheap drive, you get a cheap drive. I have friends who refuse to buy Maxtor. I have friends who refuse to buy Seagate. News flash! Same company! Cheap drives everywhere! Probably all manufactured in the same place, getting stickers based on what you're willing to purchase.
No sig for you. YOU GET NO SIG!
TidBITS system guy here. Sorry for the troubles. We had a glitch in our Apache min/max/spare/etc settings that was triggered for the first time by Slashdot traffic. (A combination of a new method to zoom images and AJAX produced a very high set of spawned children for each new visitor.)
Freelance tech journalist for the Economist, MIT Technology Review, Macworld, and others
Some simple economics play into the pricing scheme.
Consumer Equipment:
Anyone that can sell 100,000 units to customers who demand *far* less in the way of service and generalized performance can sell a cheaper widget.
Production Equipment:
**Far** fewer customers (1000 units) who demand much higher levels of service, and generalized performance demands that the drives must be way, way more expensive than the consumer product.
http://www.maxineudall.com/2010/02/should-economists-be-sued-for-malpractice.html
The people who write these articles are stupid and contribute stupidity to IT in general.
1) Apple calls these beefier models "server-class" drives; you may also see terms like "RAID edition"
There is simply no data to back this up. The vendors themselves do NOT do sufficient testing to make these claims ergo Apple can not make these claims. This parallels the so-called better failure rates of SCSI/FC 'enterprise' drives and consumer SATA drives. In the FAST paper by Schroeder we see the following quote.
". . . we observe little difference in replacement rates between SCSI, FC and SATA drives, . . . ."
2) Firmware - the closest thing to an argument here is "may prevent Server Monitor from being able to report on the drive's health"
3) Carrier - "Apple also told me that the rubber grommets that hold the drive to the ADM carrier are chosen specifically to match each drive's vibrational characteristics."
This leaves out the most important thing. "So what" - ok if the drives vibrational characteristics are not matched what happens. Is it significant? Where is the data to say so?
4) Extensive testing - Essentially arguing that Apple does burn-in testing (which you could easily do yourself) however...again from the FAST paper:
"Contrary to common and proposed models, hard drive replacement rates do not enter steady state after the first year of operation. Instead replacement rates seem to steadily increase over time."
Drives act like mechanical devices NOT electronic devices.
In general - have you EVER read an article with so many "may"'s and "possibly"'s? There's very little here that could be risk assessed (giving some kind of probability of some consequence) - which means it USELESS as advice. The parts that actually IMPLY some kind of probability/consequence are not well supported by the studies with the largest sample sizes.
Mods, this is not offtopic, anyone who read through the article as posted above would have seen the price comparisons with similar HP and Dell drives. Apple is not alone in this price gouging, if gouging it is.
Rule of Slashdot #0: You and people like you are not representative of the larger population. - A.C.
Incidentally: Our dept actually did a similar kind of upgrade on an XRAID (which was PATA but Apple had made all sort of claims: 'zero defect drives' blah, blah, blah). It's been working fine for close to a year now. The only issue was a firmware upgrade (to the XRAID) to handle the 750GB drives which would have had to be done even if we bought Apple. The cost savings was pretty significant I recall - something like 60% of the Apple price.
This was a complete replacement though. I do agree with one admins posting above though that if you are replacing a single drive then you need to be careful (and spend the money) to insure that the replacement is an exact match.
Parts of this may be true (it's impossible to say, since they don't specify exactly what they're comparing) -- but even if parts are true, it's misleading at best. In fact, some of it doesn't even seem to make sense. Let's look at real price lists from Apple and dell.
First we note that Dell doesn't seem to offer a 73 GB drive at all, so it's not entirely clear what they're comparing. The most likely possibility appears to be Apple's smallest option, a 73 GB SAS ADM ($300) to Dell's smallest, a 146 GB SAS ($349). While it's certainly true that the prices are comparable, it's also true that the Dell drive is twice as big.
For 300 GB SAS drives, the Apple site shows $650 while the Dell site shows $699. While Apple's price is lower, it's certainly not even close to $200 lower. To get a $200 price difference, it looks like they compared the full price of a 300 GB drive for the Dell to upgrade price for the Apple (i.e. the price difference for changing from the stock drive to the 300 GB drive).
For 1 TB hard drives, they have something of a point, but not a very good one. Apple's price for a 1 TB SATA drive is $450, while Dell's is $639. They fail to note, however, that Dell also lists a 1 TB SAS drive (an option not available for the XServe) for $679. Taking this into account, it looks a great deal as if Dell is simply doing their best to encourage their higher-end customers to use enterprise-class SAS drives by offering them at a purely nominal incremental cost over SATA drives.
The original article attempts to portray the situation as Apple offering prices that are at least as good as, and often better than the competition. In reality, there appears to be only one reasonable configuration where the Apple is likely to be competitive: the one using 300 GB SAS. At the low end, the Dell offers twice as big of a drive as the Apple for a purely nominal price difference. For lots of storage, the Apple offers only SATA drives where Dell offers SAS. If you're storing 1 TB of data (or more) the incremental cost of SAS is usually fully justified. There are undoubtedly exceptions, but they're not particularly common.
The universe is a figment of its own imagination.
We have two PPC xServers (dual G5s) and when the stock ADMs (3 x 250 Gb ) hit EOL, we replaced them with standard retail *server* grade drives (3 x HITACHI 1 TB's w/large cache & high MTBF). We saved a few hundred dollars and experienced no problems.
and it also tells me that there's very little novel in this tie-in between server manufacturer and storage vendor (i.e., the same company).
I guess we should commend Apple on truly grokking the spirit of a serious enterprise-grade server vendor and following in the footsteps of the giants like IBM and Sun.
Welcome to the Panopticon. Used to be a prison, now it's your home.
"The reason Apple (and every other server vendor) charges that much for drives is because that's what they want to do, and it's disingenuous for this guy to be spinning it as if Apple has something special in that regard."
It's like the whole world discovered the word "disingenous" at once. Why can't this guy be "gullible", or simply "wrong" about his conclusions?
It doesn't have to be deception. It can be poor judgment or bad research.
I've got two apple xserve servers, one old dual G4 with regular 4 ATA drives, and the new one with 3 SATA drives. I bought them both with smallest/cheapest drives, and then replaced the drives with off the shelf bigger drives. Absolutely no difference in reliability/performance. I can say that cause older Xserve drives were still running good for 5 years when I replaced them with bigger drives, and new xserve is still running good. Drives modules is no different every manufacturer has their own custom drive trays, which fit their servers only, Apple is doing nothing different here. Even our Mac's have had drives and other hardware replaced with off the shelf parts, and they work just as good, drives ironically were replaced from some Mac's cause original Apple ones failed.
I am at loss with words...
I'm pretty sure the I in raid stands for "inexpensive". Isn't the whole point of RAID to avoid paying extra for high performance disks?
I Browse at +4 Flamebait
Open Source Sysadmin
You cannot buy a bare hard drive and insert it into an Xserve, as you can with a Mac Pro (and having just added a drive to my new Mac Pro, I can say that Apple did a stunningly nice job in making it easy to add drives, especially in comparison to the awful approach they used in the Power Mac G5).
It's nice to know that Mac Pro users are finally experiencing what some of us have had for the past 15 to 20 years. As for XServes, talk about the ultimate vendor lock in...they choose the OS, they choose the hardware and you're just along for the expensive ride.
= Redundant Array of EXPENSIVE Drives.
They were right - the revolution did not get televised. It was posted on YouTube instead. All in 120 characters. SLOOSH!
And how long did the original drives last? Unless they failed in the first year, you've yet to state anything of any usefulness in the comparison other than 'I did it and it hasn't broke yet' and there is still time for it to fail. If it fails in half the time as the original drives than you've already lost money for sure considering you need to buy new drives AND factor in the cost of downtime, engineering, factoring the fixes into your maintaince window, and the actual physical person doing the work.
Once you factor in the cost of an additional failure using cheap drives its very rarely actually worth it.
Persistent Volume manager for Kubernetes - https://github.com/dwimsey/openshift-pvmanager
There's a big difference between "SATA drive in a custom, proprietary drive caddy" and "SAS drive" - not only in terms of functionality, but also in terms of quality.
In short, half the reason why SAS is of a lower capacity isn't because then they can leech you of your money for higher capacity storage - though I'm sure that's a consideration. SAS, and other industrial drives, will often not jump the gun and implement higher densities, instead preferring to use those same enhancements to increase disk reliability and speed.
~/ssh slashdot.org ssh: connect to host slashdot.org port 22: too many beers
I can understand that it may need a certain enterprise level drive to work correctly and reliably but all the marketing fluff is straight bullshit with no chaser. Drives with a good MTBF and all other goodness and magic maybe a little more expensive but not by much. The fact that it's Apple doing this makes people even more upset because last time I checked they dick you very hard over hardware upgrades on their "regular" consumer products. $1,100 for 8Gb of DDR3 1066 ram is pretty steep considering that 8Gb of the same memory will cost you about $100-$120 on NewEgg. Another interesting side note is the way that MS charges a premium for their Xbox 360 hard drives and Sony let's you use whichever one you want. It because MS blesses the hard drives through Kosher rituals and then encases it with a special chastity belt. Sony then tries to dick you by announcing a "price cut" on PS3 models based on their hard drive capacity knowing that you can swap it out with whatever size you want. Talk about disingenuous.
What a long winded justification for why apple can charge over twice as much for some "enterprise level" equipment.
I'd rather just double up on equipment for the same money.
So what is your long winded justification for why Dell and HP can charge more than Apple for some "enterprise level" equipment?
Lars T.
To the guy who modded me down from perfect to terrible Karma - Apple haters still suck
Buy their 80gb ones on their sled, from Apple. You'll take the least markup hit. Remove the drive from the sled and repurpose/sell it. (note it won't have a warranty as it is an OEM component now)
Then go buy a 1T etc sata and attach it to the sled and away you go.
This lacks the "improved" firmware for the drive, and may not be a good quality one, but this is how to save a lot of money. use that extra money to make some mirrors and protect your data in depth instead of quality.
I work for the Department of Redundancy Department.
The fanboi will believe anything some apple soda-jerk tells him, and gladly pay more for the "I'm an idiot" baseball cap with the extra special, specially-selected size strap. What an idiot!!
... how they use oxygen-free conductors in their wires to make the bits sound better.
That is all.
Him this would seem like a good location to sell some ethernet cables with gold-plated plugs.
Or should I sell some gold-plated SATA cables?
This leaves out the most important thing. "So what" - ok if the drives vibrational characteristics are not matched what happens. Is it significant? Where is the data to say so?
In high capacity drives, vibration is a huge factor. Did you not see the youtube video of someone screaming at his drives? It got posted on slashdot.
In this image : http://db.tidbits.com/tbthumbs/tn10166_System-Profiler-SAS-report.jpg
the drive model is listed as : st380815as n
2 seconds of googling shows this page : http://www.seagate.com/ww/v/index.jsp?vgnextoid=809a4d4b57cb0110VgnVCM100000f5ee0a0aRCRD
Uh... that doesn't look like a server / enterprise class disk to me. It looks like a normal old Seagate disk that Apple want to charge lots for cause it has an Apple sticker on it.
I have two xserves that I bought with the smallest drives possible. I then ripped out the 80GB drives and installed 500GB drives.
The servers have been in production for over two years without one single disk related problem.
I do the same thing with my Dell servers. The drives are cheap enough that if I have a failure, I have a stack of replacements on hand. Then I pursue a warranty replacement with the drive manufacturer.
I don't mind paying a premium for well designed hardware, but I hate paying a premium for commodity parts.
(I don't do this with my EMC SANs - they will only work with "approved" drives).
-ted
"And how long did the original drives last?"
The original drives were functioning up until we replaced the lot. The replacement was for capacity.
"you've yet to state anything of any usefulness in the comparison"
Note that this was an incidental note not a counter-argument to the idea that there is significant risk. I already addressed the alleged differences between Apple and non-apple drives in my earlier post. We did in fact add one thing to the conversation that the install was trivial and the only firmware issue was with the XRaid itself.
"there is still time for it to fail"
Clearly and even if these drives last as long as the other drives did it still wouldn't say anything meaningful (well it might cause you to doubt your idiot premise). Point being that you actually need a large sample size. Which I touch on in my earlier post.
"using cheap drives its very rarely actually worth it."
Again the largest bodies of data do not support this assertion. Enterprise drives and non-enterprise drives show little in the way of difference in terms of replacement rates. I've already posted links to the FAST study. So either there is no difference or the difference is overshadowed by other factors or this is something that only affects the very specific drives Apple is talking about.
Did you even READ the article most of the alleged differences strain credibility.
This leaves out the most important thing. "So what" - ok if the drives vibrational characteristics are not matched what happens. Is it significant? Where is the data to say so?
In high capacity drives, vibration is a huge factor. Did you not see the youtube video of someone screaming at his drives? It got posted on slashdot.
I love statements like "huge factor" - I suspect this is what I refer to as "adjectives over evidence". Your need to use an adjective like "huge" stems from your inability to actually quantify the risk.
Yes I saw the videos. I'm wondering how much you actually thought about them though. They show someone shouting at a disk array and an increase in latency as an apparent result.
But MY statement was "Do we really need special grommets to match the drives vibrational characteristics and if so what do we lose?"
Do you not see how someone shouting at a specific disk array doesn't necessarily disprove that statement (or prove your general statement)?
In case your answer is "no". Here is a handy list of things you would need to demonstrate to get to disproving my statement from yours.
i) Constant shouting causes equal and constant impact. From the video it looks like the shouting results in an big effect on latency. However it's hard to tell how long that effect would last if the shouting was constant (this is closer to the idea of using different grommets like Apple suggests).
ii) Shouting at an x-serve (or x-raid) makes a difference. See you can't assume that one data point makes the case for all data points. This is what Sacket referred to as the disastrous inadequacy of lesser evidence. For example perhaps the Sun box is so well engineered (compared to the X-Serve) that it's one of the few that actually ALLOWS this to happen.
iii) Having specific grommets can actually make a difference. Grommets beyond a base tolerance may not make a difference. Perhaps the kind of vibration that most grommets allow is not the kind that makes a difference in latency (maybe it's about frequency, maybe it's about amplitude).
iv) Once we establish that it's possible for screws to make a difference. It still remains to be proven that APPLE's grommets make a difference. Perhaps the only kind that make a measurable difference cost $1,000,000,000 grommets and are fashioned by elves or something
v) Apples grommets make a SIGNIFICANT difference. Just because screws make a measurable difference it doesn't mean its going to be one that most sysadmins care about.
4) Extensive testing - Essentially arguing that Apple does burn-in testing (which you could easily do yourself) however...again from the FAST paper:
"Contrary to common and proposed models, hard drive replacement rates do not enter steady state after the first year of operation. Instead replacement rates seem to steadily increase over time."
Drives act like mechanical devices NOT electronic devices.
Since you're such an expert - care to explain Figure 4, difference between compute nodes (left) and the file system nodes (middle) of HPC1, 1st month?
Lars T.
To the guy who modded me down from perfect to terrible Karma - Apple haters still suck
sometimes, people think they're so smart.
feel free to google for effects of vibration on drives.
Also, do get an engineering degree while you're at it.
I spelled it wrong not once, but twice.
I do know how to spell it, even if I don't demonstrate it here.
Also, do get an engineering degree while you're at it.
Is that what you have? If so thanks for making the fight against prejudging engineers as mathematically ignorant and devoid of proper experimental methodology just that much harder.
The truth is you could have five hundred billion articles on vibration affecting (in some way) drives. However that doesn't mean:
a) That constant shouting has the same effect. For example one of my datacenters is easily loud enough that one has to shout to be heard. Are the SAN units in there constantly having higher latency. Why didn't we see a benchmark difference between inside and outside?
b) That the effect is uniform across all devices.
c) That the vibration from HD's in action is somehow comparable.
d) That this can be fixed with little rubber grommets.
e) That Apples little rubber grommets can fix them.
See the thing that you are clueless on. Is that all you have is evidence that some kinds of vibration on some systems in some environments cause some kinds of problems. However the argument you are feebly attempting to counter is that Apple's little rubber grommets actually help in some significant way.
Since you're such an expert
I'm actually not but I do take a rather quantitative approach to IT. My personal experience reflects that most people take a religious approach.
care to explain Figure 4, difference between compute nodes (left) and the file system nodes (middle) of HPC1, 1st month?
Um...you indulging in selection bias? ;-)
Seriously though. Explain in reference to what exactly?
Seriously though. Explain in reference to what exactly?
TFA - and your complaint that Apple does burn-in testing.
Lars T.
To the guy who modded me down from perfect to terrible Karma - Apple haters still suck
Where did I complain that Apple does burn-in testing? That's right, nowhere.
I did complain that the article asserted Apple's doing of this as valuable and beneficial and that the largest body of data does not bear out the later.
You're still being pretty evasive about what exactly you want explained.
Are you struggling to say that when you pick a portion of a dataset post-hoc it does say there's something beneficial about burn in testing?
The fact that you claim not to understand what I want explained can mean two things: either you are stupid, or you know you are wrong and are being evasive. Figure 4 clearly shows the difference between stress testing and no stress testing.
Lars T.
To the guy who modded me down from perfect to terrible Karma - Apple haters still suck
Wow. Nice false dichotomy.
either you are stupid, or you know you are wrong and are being evasive ...or whatever your assertion is it's wrong and I've already dismissed it and as a result you are forcing me to pick amongst wrong ideas (which there are significantly more than right ideas). ...or I've already addressed it (i.e. references to selection bias and you getting to pick and choose the data post hoc) and you don't yet understand why that is wrong.
There could be other options but all I need is one to show that your argument about me being "either X or Y" as false.
So am I to infer here that you are referring to the fact that at 1 month, the AFR reaches 6% on HPC 1's compute nodes in the first month and are implying that this says something generally about stress testing utility?
Just say if this is the case or not...and I'll explain the problem with that - no offense but considering how evasive (and illogical) you are being. I'd like to actually pin your argument down before I destroy it.
See, that's the difference between you and I. You see (or rather, hear) a noisy data center, and assumes that it's the same inside the actual chasis. Whereas, I believe (since I've not actually researched this, and only know about it second hand) your SAN units have actually have vibration dampeners designed in.
So, it's already different.
But that's OK - if it makes you sleep any better, go ahead and say Apple charges 130% extra for their special rubber grommets.
So what is the difference between the compute and the file system nodes - no more weaseling around.
Lars T.
To the guy who modded me down from perfect to terrible Karma - Apple haters still suck
Again you have avoided a simple and incredibly direct question. All you needed was to say yes or no but you couldn't. If you state your argument clearly:
i.e. Burn-in testing is useful for most people because it would catch at least one in twenty drive failures this is based on the 6% AFR figure for one month.
Then I could get around to actually showing you why you are wrong. When you keep beating around the bush and then make (ill-informed and illogical) accusations when someone asks you what you actually need explained. It's going to be difficult showing you where you are wrong. However I expect that's what you want.
I've already stated what I believe but here it is again for the record: Apples burn-in testing does not add significantly to failure prevention compared to drives purchased over-the-counter. I expect that we are looking at something close to a failure detection rate of no better than an average risk of 1 in 300 to something worse that 1 in 500. Which is hardly a significant risk in buying a single drive for an x-serve or a set of drives for an Xraid (conflating both my example and the articles examples). Furthermore even if that is a risk, it's only a service of value to those way outside of the articles target audience.
So what is the difference between the compute and the file system nodes - no more weaseling around.
Again, I'm stuck decrypting your statements into an actual argument. So what do you want here? Personally I see little that can be definitively stated as a difference between the systems (the type of drive use could be inferred). That said there is little reason to believe that the failures shown in the first month of compute node usage are important to a *general* - This is an important word - argument which is what you appear to be making.
You keep digging yourself in deeper and deeper. I'm not interested in your trolling. Go away.
Lars T.
To the guy who modded me down from perfect to terrible Karma - Apple haters still suck
See, that's the difference between you and I. You see (or rather, hear) a noisy data center, and assumes that it's the same inside the actual chasis.
Don't mistake my beliefs with outlining the assumptions you need to make your religious system work.
Whereas, I believe (since I've not actually researched this, and only know about it second hand) your SAN units have actually have vibration dampeners designed in.
That's actually irrelevant. So you are confused about something but it's unclear exactly what.
go ahead and say Apple charges 130% extra for their special rubber grommets.
That's also irrelevant to the discussion. Apple is charging for a lot of things (branding for example) the article is talking about the comparative value of taking an ADM and then stripping out the drive and replacing it with a non-Apple drive. So the only vibrational difference outlined here is "the rubber grommets that hold the drive to the ADM carrier" and the reason give is that they "are chosen specifically to match each drive's vibrational characteristics"
So I asked the all important question "So what" about "if the drives vibrational characteristics are not matched" (via these special grommets)
And you made some obtuse reference to the JBOD / Dtrace clip that's been going around everywhere.
Ahh...and you lose.
All you had to do was state your argument clearly and then it could be discussed. However you avoided every opportunity to do so. If you had a cogent argument (or a disproof of mine) one would assume that you would have stated it by now.
On the other had I stated my position at length AND answered your most recent question to the best of my knowledge.
So how does that make me evasive?
I have been replacing drives inside the adm's for years to avoid the unjustified high pricing of apple's adm's. Not only is this less expensive but you also get a much better warranty. This article is mostly nonsense.