But they won't guarentee it for >3 years

Post #32,930 by tonytib 3/20/02 8:01:20 PM Reply	But they won't guarentee it for >3 years And, I'd much rather the HDD not fail at all than have the manfacturer replace it, with all the trouble that causes. Some questions: -- What is the realistic 24x7 life of an IDE hard drive? SCSI hard drive? -- How much does it vary by type (IDE vs SCSI) and manufacturer? I doubt if we'll ever really know the answers to these questions. But I doubt that IBM's IDE hard drives (except for 75GXP) are really any worse than WD, Maxtor, etc. OTOH, I can believe that SCSI drives will last better than IDE in server or RAID use. If I'm going to spend the money on a RAID array, I'd like to last > 3 years. Based on the experiences of Karsten and Andrew, I'd seriously consider spending the extra money on SCSI RAID. Tony
Post #32,936 by a6l6e6x 3/20/02 8:57:25 PM Reply	No disagreement from here. Alex "Never express yourself more clearly than you think." -- Neils Bohr (1885-1962)
Post #32,957 by Andrew Grygus 3/21/02 12:17:57 AM Reply	Highly erratic Some models of some brands are all dead and making klunking noises inside of three years, while their predecessors and successors are still happily running five or ten years later. Based on my averages I avoid Seagate, but others brands have had some dud models now and then. Individual drives from a good production run are also unpredictable. What I tell my clients is: "On average they probably run between 5 and 10 years, but that doesn't mean your hard disk isn't going to fail in 10 minutes". I haven't seen a head crash for quite a few years, and shaft chatter isn't anything like the problem it once was, nor is sticky heads. Most of the failures I see are sudden and catostrophic (no data recovery possible outside a $ cleanroom data recovery house $). Most common is failed head circuits (can't find track 0). Commonly this is not on the main logic board, because swapping the board doesn't fix the problem. Second most common is motor electronics (doesn't spin). Broken positioning mechanisms were a problem a few years ago, but I haven't seen that lately. Following far behind is failed video cards (they can take just about everything else with them) and then a bit of everything. Most common client comments are: "Backup? yes, we have one from about 8 months ago, but I'm not sure I can find it. Can't we get it off the hard disk some how?" "But it worked just fine all day yesterday. Why would it just stop working?" "I was just saving a file and this error message popped up. How could it just do that?" "I thought this was a good brand?" "Three thousand dollars to get the data off? Well, my son says he knows somebody who's and expert and will do it for free." [link\|http://www.aaxnet.com\|AAx]
Post #32,993 by tonytib 3/21/02 11:28:38 AM Reply	I especially love that last comment!
Post #33,084 by folkert 3/21/02 9:15:46 PM Reply	Interesting.... Very interesting... [RANT] not just at Tony Funny comments about the difference betwixt SCSI and IDE. Have any of you ever really been a a HD manufacturing facility??? Nope, is probably the answer. You ever notice anything similar about Western Digital SCSI and IDE drives... How about Seagate's??? Maybe we should talk about Quantum or even IBM! I want you all to honestly think about your comments you all have made here, think really hard... then see what has happened here. I guess you all will not agree with me, BUT the biggest difference between MOST and I mean MOST SCSI and IDE drives is the endcoding on the media, and the electronics on the drive. People will always say SCSI is better... mainly it is only FASTER for sustained through put and more flexible in configurations. Ever see a build sheet for SCSI and IDE drives... Same part numbers except for the drive electronics and for the higher rotational speed drive the pan motor and head actuator (typically more dampening for fast spinners). OKAY. now that that is over... Granted IDE is more "stand-alone" oriented, Granted we tend to expect the SCSI flexibility, Granted we tend to assume that IDE is "inferior", Granted we all like to gnash our teeth on "esoteric items" failing at critical times... BUT... IDE drives have become more ubiquitous and more plentiful than maggots at a garbage dump. IDE drives are currently being used in about 60% of high-availability NAS solutions... even IBM uses them in some versions... EMC is currently determining where it needs to be in that arena. RAIDZONE bases it WHOLE business model on IDE drives. Storage Appliances mainly use IDE drives for storage. Most all high-availability solutions use some kind of RAID + hot-spares configured into them. There in lies the "trick". That single trick allieviates 99% of those issues you are complaining about. Sure you still have the problem of fixing the failure, but you are not running in degraded mode and can breathe a bit. Sure, I have had my share of hot-spare not swapping in automagically(recently an IBM IPSRAID card failed to do that for me). Storage is not something to take lightly, if you want it to last and be available ALL THE TIME. You have to make sure you plan the redundancy in the system. Don't just hope it'll live. Sorry, bitching about failed arrays that have no redundancy in them, be it failed controllers or failed drives or what have you... Done. got it out of my system. (pun intended) `greg, curley95@attbi.com -- REMEMBER ED CURRY!!!` In 2002, everyone will discover that everyone else is using linux. Linux: Good, fast AND cheap. Failure is not an option: It comes bundled with Windows. ** "Two rules to success in life: 1. Don't tell people everything you know." - Sassan Tat
Post #33,090 by tonytib 3/21/02 9:46:09 PM Reply	Greg, I've been looking at actuators for six years now plus I've done work for HDD media companies. So I've seen plenty of SCSI parts and IDE parts, mainly actuators and head stacks. And, I can tell you that the basic quality of a Seagate SCSI head stack is better than a typical WD IDE one. That doesn't mean the whole drive is better, since that depends on a lot of factors. Some SCSI and IDE drives only differ interface electronics, but many do not. For example, there is no IDE equivalent to the 10,000 RPM SCSI drives (Fujitsu, IBM, Seagate), let alone the Cheetah X15 (15,000 RPM drive). And, running at higher speed requires quite a few changes. For example, high-end SCSI drives typically have lower areal density than IDE drives. Frankly, I'd like to see solutions that just work. I'm sick and tired of messing with stupid computer components that are a pain to get working. If (or when) I need a RAID (level 5) array, if I can be confident that the SCSI solution will be easy to get working (and be more reliable) than an IDE RAID solution, then I'll definitely consider spending the extra money. Storage Applicance, EMC, etc can spend a lot of time and money figuring out weird permutations in IDE RAID configurations (if Karsten and Andrew's experiences are a guide). I don't want to - and I bet I can build a Linux-based NAS solution with SCSI RAID drives that will work fine for me for a heck of lot less than them -- or even Quantum (SnapServer), etc. Tony
Post #33,100 by folkert 3/21/02 11:05:09 PM Reply	Then why do we never here about... SSA... huh? Current revisions of SSA is extremely fast, Extremely reliable... matter of fact IBM uses them in thier Enterprise Storage Server (ESS - Shark). Lets assume that there are significant reasons why IBM bucks the trends. Lets grant that IBM is all knowledgeable. Lets just wonder WHY SSA isn't as popular in PC servers as they are in RS/6000s. In my extremely NON-Humble Opinion, SSA drives with SSA Raid controllers are Far and Wide the most reliable, fastest, scalable, easy to service, easy to make work, self diagnosing, points you to the problem, SETUP in all the world for Storage. I can remove half the wires connecting the drives... as long as 1 remains connected between all of them SOMEHOW, they are all still available. They have zero degradation in performance in that situation, SSA RAID uses can make SCSI raid pale in comparison (even the mighty Mylex controllers IBM makes do too) So explain to me people choose SCSI over SSA. Based upon you arguements SSA would be the ultimate.... PERIOD. I'll tell you why... everyone has been brainwashed into thinking SCSI is BEST, IDE SUCKS anything else is "esoteric". SSA can sometimes be as much as 200%-600% more expensive than even the fastest SCSI. I have gotten IDE RAID to work as well as SCSI RAID, including failure handling. Some may indicate 3ware having problems, company has some developments coming down the pike that may surprise some of us. They are going to be gaining ground. Now I agree earlier products from 3ware were/are buggy. But they have updated firmware for them... but component failure can't be fixed... Oh well the question remains: Why is SSA not as widely used as it should be? `greg, curley95@attbi.com -- REMEMBER ED CURRY!!!` In 2002, everyone will discover that everyone else is using linux. Linux: Good, fast AND cheap. Failure is not an option: It comes bundled with Windows. ** "Two rules to success in life: 1. Don't tell people everything you know." - Sassan Tat
Post #33,105 by Andrew Grygus 3/22/02 12:06:45 AM Reply	Could be a bunch of reasons: Ownership of IP (Intellectual Property) may restrict the number of vendors. Cost of licensing IP may make alternative vendors uncompetitive, so they don't enter the market. Single source is not wise in most (but not all) cases. Cost may be beyond the point of diminishing returns for the majority of applications, just as the cost of SCSI is beyond the point of diminishing returns for most desktop PCs, but priced just right for midrange servers. Lateness to market may make marketing cost to achieve economical volume too high, limiting the product to only the highest end market. Products are designed for particular markets. IDE was designed very deliberately to reduce cost in volume production. Put the controller on the device, and make it a simple controller. SCSI was designed for greater flexibility and better performance. SSA was designed for very high end systems. You can put 2 IDE devices on a cable, 16 SCSI devices. Even with just two, IDE devices don't share the cable very well. Hell, when I put a fast CD writer and a CD-ROM in the same machine I have to put them on separate IDE cables or I get underruns making copies, and CDs aren't really all that fast. I don't have that problem with SCSI. True, newer IDE drives outperform older SCSI drives (at least if there's only one IDE drive on the cable), but SCSI is a moving target. Soon we will have Serial IDE which will solve the "2 per cable" limit, allow longer cables and greatly improve data transfer speed, but Serial SCSI has already been in development for quite some time and will be there for the more demanding markets. Right now, high performance IDE is operating system dependent because of the special drivers needed. SCSI isn't. You can get high performance out of SCSI even on OS/2. I seriously doubt I could get SSA to work at all with the small business servers I build. True, many SCSI drives are mechanically the same as equivalent IDE drives except for the control circuitry, but many aren't. Different SCSI drives are targeted to different markets. Right now IDE doesn't have that differentiation - it's all aimed at the lowest price market. Trying to force any of these devices into the others' markets is not going to work well. [link\|http://www.aaxnet.com\|AAx]
Post #33,110 by tonytib 3/22/02 1:18:29 AM Reply	Politics, IIRC SSA was IBM's equivalent to FC-AL, and it lost out during standardization wars, IIRC. So a number of people use FC-AL, notably Sun, and FC-AL drives frequently appear on the surplus market (I don't believe I've ever seen an SSA drive) and even at Fry's occasionally, but only IBM uses SSA. I don't know enough to compare FC-AL and SSA, so don't ask. My internal knowledge of hard drive parts is based on the parts I've seen, mainly HSA's. Thus, my quality comments were based on the mechanical quality of the parts I've seen (view from inside), not on interface abilities, reliability, etc. Although I've picked up a fair amount of general knowledge (e.g. from IDEMA magazine, DiskCon, customers, etc), I'm not a hard drive designer. OTOH, I'm pretty sure, for example, that the Seagate SCSI engineers don't know much about the Seagate IDE drives (with the possible exception of the IDE Barracuda models if they share the same hardware). My interests in RAID are unrelated to my work. I'm interested in RAID 5 controllers for home (maybe if the economy picks up) and small business. SSA and FC-AL are too pricey. IDE is a possibility, but I'll have to see a lot of good comments (e.g. on StorageReview.com) before I'll trust 3-ware. Promise and Adaptec are also possibilities. As I said before, I'm tired of PC stuff that half works. BTW, the Mylex SCSI RAID boards appear to be about the same price as the 3-ware IDE controllers. As far as future greatness goes, well, my standard answer is: "I'll believe it when I can see it at Fry's". A lot of products (not just from MS) never make it out of the vapor (Linear Technologies, for example, had an analog chip that they had spec sheets for, advertised, and engineers started to design with -- but the chip never made it to production. So, a common EE rule is "If it's not in the DigiKey catalog, don't use it.") Tony
Post #33,889 by Brandioch 3/30/02 10:45:06 PM Reply	Then what makes SCSI seem better? I'll echo Andrew's comments about CD burners. EIDE >REQUIRE< that they be on different controllers. I also SEEM to have better response/performance from SCSI hard drives in servers. If the components are the same, is it the controller that makes the difference? Not that I'm adverse to putting a dozen or so controllers in a box and running IDE.
Post #33,892 by folkert 3/30/02 11:22:21 PM Reply	The what! See [link\|http://z.iwethey.org/forums/render/content/show?contentid=33848\|Remodeling gone awry]. Just remember, IDE sucks with more than one device sharing the channel. That is why 3ware uses 4,8 and (soon to be released)12 IDE ports on a single card. The card support only ONE IDE drive per channel. So you say this is a down fall, well SCSI signalling is different, and able to handle multiple different speed devices on the same BUS. IDE can't. Just is that way. If you look at the link above, you see what I did for a company. Early numbers maybe indicating the only thing the SCSI setup has over the IDE setup, is better heat generation. As of right now this minute the transfers on the IDE "Logical Drive" are coming in at about 600-610MB/Sec, but I believe that is limited cause of the memory testing going on concurrently, along with a couple of kernel compiles at the same time. The SCSI "Logical Drive" is currently averaging about 550-780MB/sec depending on stuff staying in the (128MB) controller cache, I believe. Gonna switch over to a low-latency kernel and see what happens, some time tomorrow(easter). `greg, curley95@attbi.com -- REMEMBER ED CURRY!!! Keep software free. Oppose the CBDTPA. Kill S.2048 dead. [link\|http://www.eff.org/alerts/20020322_eff_cbdtpa_alert.html\|[link\|http://www.eff.org/alerts/20020322_eff_cbdtpa_alert.html\|http://www.eff.org/...a_alert.html]]`
Post #33,932 by Brandioch 3/31/02 3:24:08 PM Reply	Sounds good. The only things I need are: #1. It has to be as FAST as the SCSI stuff I work with today. #2. It has to do RAID 5. (preferably two RAID 5 arrays setup in a mirrored configuration (RAID 5+1)). #3. The drives have to be hot swappable. #4. It does all the above in hardware so the OS doesn't see anything but one big drive that it can boot off of it. #5. Lots of blinken lites. We've used a Snap! server before. Four drives (I think) in a RAID 5 array. It wasn't bad, but it was SLOW to back it up with ArcServe. The NetWare and Windows boxes would zip right along. This one would CRAWL at a few megs a minute. Then, one day, it just decided to forget its array configuration and lose all the data. The speed was the problem for us. It was okay writing to, but a dog getting the stuff off in mass quantities. I'm sure there was a configration setting that would solve this. Whether it was available to us end-users is another question. Now, I'm thinking about turning it into an MP3 server. Just to hammer on it and see if I can make it fail again. I'm going to have to try the 3ware stuff in my toy servers soon. Thanks.
Post #33,934 by folkert 3/31/02 4:27:47 PM Reply	Well, it all depends... #1. It has to be as FAST as the SCSI stuff I work with today. You saw my numbers so far... They are still holding out. Gotcha covered there. #2. It has to do RAID 5. (preferably two RAID 5 arrays setup in a mirrored configuration (RAID 5+1)). It does do RAID-0 RAID-1 RAID-5 RAID-10 with Hot-Spare. Gotcha sort of covered here too. #3. The drives have to be hot swappable. IDE Hot-swap cages are coming out of the wood work these days. The one 3Ware actually sells works very well, but one has 3 slots per cage (1 full-height slot). The cards support hot-swap as well. Gotcha under the blanket now too. #4. It does all the above in hardware so the OS doesn't see anything but one big drive that it can boot off of it. Offers JBOD, RAID0, RAID1, RAID5, RAID10 and I think Linear array all behind the smoke and mirrors so the OS just addresses one big dumb Drive, and can boot from it. Blanket seems to cover that. #5. Lots of blinken lites. 1 Blinking light per IDE channel, plus 4 for status... just on the controller itself. Now as for thier H/S IDE cage, 2 LED per drive, one activity, one powered/status(tri-mode). Now hows that for lights. Been sleeping comfortably now I see. Now the reason, I setup my things the way I did, I like to be able to replace the "system" in case of a catastrophic failure on the machine. Then just move the controller/subsystem to a "new" machine. The only question I have for you, What do you really need RAID5+1 for??? Speed? or diaster recovery/redundancy? Speed, ain't no big thing. Remember this IDE card only has 1 drive per channel. (ATA-133 theoretically gives you 133MBytes/sec * 8 Channels) SCSI has multiple per channel. You only get 160MBytes once per channel. Diaster recovery/Redundancy? Well, here is basically what I am going to say. Unless you have Fiber Storage at least 2KM from your primary Data you are hosed anyways. Setup you raid-5 arrays with Hot-Spare(s). Even setup you Mirrors with Hot-Spares. Setup all your fault tolerant arrays with hot spares, and alot of your concerns about SCSI or IDE go away. Oh, one more thing(I hope just 1 more), backing up a SNAP! server or ANY target without a TSA is a bad thing. Make things crawl period with ANY backup solution. Even Tivoli Storage Manage(TSM) (formerly Ad-Star Storage Manager or ADSM) doesna worky well without a Target Service Agent on the machine the Server itself runs on. `greg, curley95@attbi.com -- REMEMBER ED CURRY!!! Keep software free. Oppose the CBDTPA. Kill S.2048 dead. [link\|http://www.eff.org/alerts/20020322_eff_cbdtpa_alert.html\|[link\|http://www.eff.org/alerts/20020322_eff_cbdtpa_alert.html\|http://www.eff.org/...a_alert.html]]`
Post #33,954 by Ashton 3/31/02 10:26:14 PM Reply	Hey there, Greg It seems as if that Tek 454A went to a good place - it's still the best for lo-level (2 mV/div!), super-sharp trace sleuthing at the 150 MHz level = er reduced BW at That low a level. Now whatcha need for the above fancy stuff! at near 600 MHz BW* and whatever 7000-series plug-ins turn you on (like mV at 100 MHz differential input? check out those r/w heads Directly..): * "400 MHz nom." measured = 600; now That's conservative.. is a Tek 7854, 4-plug-in mainframe. I gots an extra! It provides pure analog == Real, not that digital-approximation stuff BUT: also digital storage and manipulation of the data. This can be saved indefinitely via just a simple PS keeping the mem. alive with scope off. Iterate n-times and you can see the glitches too. And the number of illuminated PBs and fancy delay switching lights.. warm cockles of heart. Yeh it's bigger, but it's beautiful and has a keyboard for programming some rather clever automatic tests: the honesty of analog + digital 'convenience'. (I was checking rise-time with a 50 pSec generator: clear sharp traces even at 0.5 nSec/DIV, unmagnified) Anyway, the world is crazy enough that, such lab grade equipment is now ridiculously cheap. Til all the good ones are gone. 'Course you may be too busy using this stuff to test it ;-) Ashton
Post #33,958 by a6l6e6x 3/31/02 10:44:51 PM Reply	OT: Tek scopes Ashton, you are terrible. You had to remind me of a day of joy when a Tek 545-B with dual trace and delayed sweep arrived for my use. What a quality work horse that was! Yes, I know it's a different scope, but the model numbers are a hair away in memory. Alex "Never express yourself more clearly than you think." -- Neils Bohr (1885-1962)
Post #34,077 by Ashton 4/1/02 8:18:26 PM Reply	Yeah.. imagine finding Art in a lab. I remember too, when the first 545 [no -B] arrived at the electron synchrotron - replacing a 514 AD IIRC (?) All those ceramic terminal strips, anodised aluminium - and as close to perfection as a manufactured product might get. We just stood around looking at it, with the covers off. (What a great living room display for the Upscale folk - even today - leaving off the covers! Especially late at night = foot warmer too) As doubtless you are aware, quite a few are still around and in (or restored to) original condition. Pity that 500ish watts is just a bit much and 30ish MHz a bit little :( But with sampling plug-ins, BW is still respectable; not a few 1L4 (?) spectrum analyzers are still being used too. I have a 564B storage scope with a Nelson-Ross audio spectrum analyzer plug-in. Gotta find an audio type to get this one. Clean as new - and the -B mainframe was all solid state (though the Ross unit ain't.) Anyway.. Tek ain't Tek anymore - just as HP has become a commodity reseller of $30 ink cartridges at 500% profit. Still, I've culled pristine examples of the clever Sony-Tek small scopes, like the small 35 MHz 335; another model of same size at 5 MHz / w DVM, etc. These peddled for >$3k towards the end of their run == now just hundreds, and some like new. Someday soon, people won't believe an actual "Corporation" was ever motivated to build such quality. {sigh} Quality - look for that in your local museum. Ashton hoping sometime to steal a 2467-B "Bright-eye" scope, just for the hell of its impossible complexity / simple operation.
Post #34,088 by a6l6e6x 4/1/02 10:51:18 PM Reply	Re: Yeah.. imagine finding Art in a lab. Did you ever note the little cartoons on the schematics in the manual? Or, the small roll of silver solder inside the case for an unlikely repair? Alex "Never express yourself more clearly than you think." -- Neils Bohr (1885-1962)
Post #34,114 by Ashton 4/2/02 12:39:05 AM Reply	Yup.. Loved the skier.. down a delay-line IIRC !! The little rolls of, technically "3% silver-bearing", not really silver-solder - so as not to cause the 'fritted silver' on the strips to come loose with repeated tin-lead soldering. Perfectionist stuff, and on its own white ceramic-like bobbin! But what the New-Tek is like: I burned up the First e- thing ever (well, almost) on a #$#@% 2430 digital scope: seems that it May Not be operated out of the case w/o "an additional air supply". Now there were labels of usual dummy kind.. CAUTION Glass CRT under high vacuum* and the like.. but only (later when it came) in the manual ~ p.45 was a little note about this peculiarity. And NO LABEL inside the case. D'Oh. Really pissed me off because I was able to cure an error message about a particular part by forcing a re-cal; scope would have been OK. But then I recalled.. in space, nobody can hear you scream.. Grumble, piss, moan
Post #33,994 by folkert 4/1/02 12:14:08 PM Reply	Hey there, Ashton Just to let you know, the 454 is firmly mounted in my workbench I finally got around to making. It's where I make all my junk come to life.... MUHAHAHAHA!!!! Believe it or not, I haven't even come close to running out of BW on the 454A. It works just fine. I actually had to replace a couple of Electrolytics in the Power Supply last month... It was acting funny... perfectly good one minute... wandering all over the place the next. So, I hooked up my trusty "self-made" fet VOM I built from scratch (using Plywood as the face) and saw voltage rising and dropping...rising and dropping right after the filters on the PS. Started looking at the components to see if they all had thier magic smoke in them still, sure enough 2 Caps had split covers. Apparently boiled at some point not to long ago. But put expensive replacements in... voila she good as new, No? BTW, you should really look at the new "synthetic" Electrolyte companies are using in "premium" Caps. It smells terrible, even compared to arsenic based electrolyte. But, it is black and gooey. Supposedly has 5-10 times the insulating and cooling capability of standard electrolyte. Same size caps have high capacitance and higher volage ratings. L8RZ. `greg, curley95@attbi.com -- REMEMBER ED CURRY!!! Keep software free. Oppose the CBDTPA. Kill S.2048 dead. [link\|http://www.eff.org/alerts/20020322_eff_cbdtpa_alert.html\|[link\|http://www.eff.org/alerts/20020322_eff_cbdtpa_alert.html\|http://www.eff.org/...a_alert.html]]`
Post #34,072 by Ashton 4/1/02 7:56:30 PM Reply	Good sleuthing and thanks for the tip on someone finally looking for insulation AND heat transfer (D'Oh..) I bought an Aussie designed clever (and small) ESR measuring device for that service (of course when the thing is split.. well, that's lucky). 2 LEDs read out the ESR; a clever IC-design, I thought. As you know, an ohms check will tell shorted, but a bridge won't know about the series ohms. Anyway, have a few (!) 485s and I don't want to have to unsolder / substitute all those barely accessible thingies. The meter sits there waiting for me to attack the pile :( Wanna fix some switching supplies accompanied by rest of 485? As to 454 BW being adequate.. I thought you wanted to find Truth, not just an approximation! :-\ufffd

Welcome to IWETHEY!