Seagates are usually (in my recent experience) about $5 to $15 more for a 250 GB drive. I find that Hitachi and Maxtor drives are usually very close to the same price as WDs for comparable feature sets.

Seagates are certainly worth the cost, but you may be able to find another good brand with a price the same or even lower than a comparable WD drive. That's why I'm running Hitachis, for instance.

... yet if you go to CompUSA, you regularly find cheaper comparable Maxtor drives?

Personally, I use Western Digital exclusively (since all my Maxtors (of different models and families) died), and have yet to have a drive go bad (whereas literally every Maxtor drive I've ever seen has gone bad within ~3 years).

But that's mostly luck.

In reality?

I don't know. Ask 10 computer experts, get 11-12 different answers.

www.storagereview.com's reliability survey (the only fairly large one that I know of) seems to show that Maxtor is less reliable on average, so I'd avoid them. Plus, they're often the cheapest around — kind of a bad sign.

A lot of people will have buy a batch of drives from the same manufacturer and have them go bad, and think that manufacturer produces bad drives — which is misleading. Every manufacturer has bad batches, and every one has bad drives, too.

You really want reliability? Get SCSI drives. They're much more expensive, but because they're destined for servers, they're much, much more reliable.

Otherwise, Hitachi/IBM, or Seagate, or Western Digital. Specifically, I'd avoid Maxtor (because of horror stories and the referenced survey) and Quantum (just from so many horror stories).

The best advice?

Backups are not optional.

I also use Western Digital Caviars exclusively, and always have.
Right now I've got two of these in my computer at home.
I've never had any trouble with WD hard drives.
But it's all a matter of opinion when it comes to hard drives,
as with anything else inside a computer, I suppose.

Well, You aren't going to be able to buy a Maxtor soon, as they have just been bought by Seagate. No great loss to me as I have never bought a Maxtor. Does Quantum still exist? (It seems they do, although they seem to be out of the consumer market.) I thought they went the way of Micropolis (which I also had a high failure rate on their 9G 5 1/4 full height drives. From a long time ago.)

SCSI drives do tend to have MUCH higher MTBF, but again, they tend to be Enterprise class drives meant for 24/7 server use. However, you pay for the interface and the warranty they come with. The PC has tended to go with the SATA, as it has taken over the desktop.

Backups aren't optional, but they are getting where they aren't practical. Much like Scor discussed, I went with a storage box, with mine running RAID5. It allows you to lose a drive without losing data. Had I gone with RAID 5 + HS (an option on this box), I could have lost 2 drives with no data loss, with the compromise of only having half the useable space, as apposed to 75%, as I have now.

Of course, if you only have a small subset of CRITICAL data (700M or 4.3GB), CD or DVD backups are still practical. But still anoying enough that most don't do it on a regular basis.

Over 90% of the computer pros I've talked to (over 100 in the past few years) have been pretty consistent on the subject of HDD reliability by brand, which is why my position is what it is now.

But you're right, there can be lot of variance too, depending on who you talk to. The ones I put the most creedence in are infrastructure managers. They're the ones who buy all the technology hardware and software for the company, and have to replace components when they die. And they don't like buying unreliable stuff, because they buy components by the hundreds.

I'll give you Maxtor if you give me WD. ;)

But seriously, I wouldn't argue too much for the side of Maxtor. I've had good luck with them, and most others I've talked to have as well. But we know they're not as good as Seagate and Hitachi. I'd agree with you easily that Seagate and Hitachi are very good buying decisions.

Amen. Regardless of the brand of your hard drive(s).

What I do is use two external 400GB drives, one online and one as an offline backup.

I backup all critical data from the local drive of each of my PCs to the online 400GB drive at least once a week.

Then once a month I plug in the backup drive and do a full backup of the online 400GB disk volume. When the backup is done, the backup drive gets disconnected and put into a fireproof lockbox.

My most critical - and most active - things also get backed up to multiple other locations (jump drive, CD-RW, DVD+RW, webhost, etc.) on the fly (as they're created/updated or as needed).

My recommendation for a PC or network storage strategy:

Buy and install the best drives you can afford, but backup as if you've bought the cheapest, worst hard drives in existence.

WD has been ok for me so far - although I think there's ony 2-4 total, whereas most of the computers have maxtor. The portables are iOmega. So far so good with all.

Is iOmega the brand of hard drive, or just the company that puts one into portable casing?

I'm pretty sure Iomega subcontracts with a hard drive manufacturer for the actual disk drives. But I'm not positive.

I think I may go for a storage array on a separate box. Since I'm building a new box anyway. I can just get a smaller hard drive for the box itself and put any serious money i would spend on hard drives into a raid array that I can use for all the computers in my network (dorm room)

As if that isn't sad enough, I am working on a website that I plan to host myself on my old box and I'm seriously considering running my own dns and mail servers.

The reason for all of this is that my future job is most likely in netowkring and how great would it look on an application to say that I run my own main and DNS servers already (not to mention my current job in the school as general network person (I fix stuff when it breaks, terminate cables, write scripts, etc etc)).

These are getting more popular and more economical. I'm doing this next time, myself.

Except I'm not going to connect it to a PC, but to my router as a networked device. I never know which machine I'm going to be on, and I don't want to have to worry about which is connected and sharing at any given moment. Essentially a dedicated SAN. Probably RAID-5, with around 2 TB of usable storage capacity.

One of my PCs will always remain at home as a file server. At some point I'll also configure it as a web server, so I can share my files, music and other media over the internet from anywhere in the world. That way I can take this (notebook) PC with me, and not have to carry around an external hard drive as well. Its local HDD will only contain installed apps.

My goal is transparent, turnkey remote access to my media and information from any PC in the world.

And eventually from my cellphone. And then later from my satellite-linked, wet-wired cranial implant.

:)

That's great if you have fast internet access wherever you go.

I'm stuck going places that have either cheap-ass dialup or no access at all.

Therefore I usually have to tote the laptop, portable hardie, and all sorts of wiring for this reason.

I wish I could do what you're proposing, but not enough places I end up at have the broadband that I would need to stream files.

That's a good point. But then again, I'm not going to do this for a year or two. Broadband is becoming more and more widespread. I'll only do this when over 90% of the places I go - and 100% of the places in which I'll have a critical need for my files - have reliable broadband access.

Then again, in a few years I'll have at least a terabyte on my keychain flash drive, so this might all be moot.

Well, I haven't spoken to that many IT people personally, so maybe you're onto something.

Personally, I believe that all hard drive manufacturers know how to build a reliable hard drive — doing it cheaply is another story entirely, which is why I think SCSI is probably the way to go. As a step down from SCSI, though:

Western Digital RE2. Sexy. And comes with a five year warranty. They're just not going to be offering a five year warranty unless the drive is pretty damn reliable. Seagate has a similar offering, their NL35 series. Also comes with a five year warranty (but then, so do all the other Seagate drives I saw... hmm... I'm liking Seagate more now :D)

Not neccesarily. Just because X has a Y year warranty, doesn't mean it will be trouble-free for Y Years. It just MAY mean you shouldn't have to pay to have it fixed for Y years. When the Korean cars first came out with a 7 year warranty, would you trust a new one on a cross country trip? I wouldn't have. The situation may be diffent now (particularly Hundai); I cannot verify.

Just because they upped there warranty doesn't mean JACK about their reliability. They are two completely different things.

Seagates for me have been trouble-free. WD's have not. YMMV.

Even if they did, what they "know how to do" is irrelevant to me. What they do is all the matters. And CM records are the measure of that.

Oh, you'd be very surprised. ;) Additional sale revenues gained from advertising the warranty cover an increased number of product replacements, to an acceptable threshold. This is how warranty durations are determined.

1.2 million hours MTBF on that WD drive...in "non-WD runtime hours" that's about 60,000 - around 7 years in a perfect environment*. As long as the buyer knows this and plans accordingly, then they'll probably be OK. But with WD that's not as sure of a thing as with other brands. It's just a matter of how much you're willing to gamble with your data.



* Many infrastructure managers I've worked with or managed have used specific factors when determining the actual MTBF of certain drives from certain manufacturers. A factor of 0.05, for instance, has been used when calculating the actual MTBF of Western Digital drives (there are some other brands with this factor as well), across all interfaces.

Seagate or Hitachi drives, by comparison, would provide an actual MTBF using a factor closer to 0.075. 1.2 million hours MTBF = 90,000 actual hours MTBF (>10 years). (Seagate has specifically rated this factor, and I'd put Hitachi along side it.)

I've taken the word of hardware managers regarding these adjustment factors (and sometimes blindly), but in retrospect they DO accurately reflect over 20 years of actual experience I've had with disk-based magnetic media technologies, and vastly more years than that by the people I've learned from.

You're ignoring the fact that different drive manufacturors make different lines.

No I wouldn't. If even, say, 30% of drives are going to fail within 5 years, they're simply not going to make money with a 5 year warranty. You can't make up in volume with a negative margin!

I don't think that's true. But neither of us really has any hard data, do we?

Using your rating scale:

Referenced WD Drive Series: 1.2 Million Hour MTBF * 0.05 = 60,000 Hours (~6.9 years)
Referenced Seagate Drive Series: 1 Million Hour MTBF * 0.075 = 75,000 Hours (~8.5 years)
Seagate SCSI Drives? 1.4 Million Hour MTBF * 0.075 = 105,000 Hours (~12 years)

Best solution - back everything up. Yes, it takes twice as much memory, but your data is more than worth it, especially since storage devices are down near fifty cents a gig or less. I have at least original plus two backups of everything. I could lose my job if I lost my data.

Use Norton Ghost, or if you hate Norton, as I do, use something similar.

I just use Norton Ghost beacuse it was free with puchases of iomega portable hard drives. It's the only Norton product that hasn't yet pissed me the fuck off (excuse my language - it's just that other Norton products have caused me some agonizing problems.

Not true. Manufacturers use different MTBF ratings for different lines and different interfaces (EIDE/SATA/SCSI). So the same factor for works most of the time for a particular manufacturer. There are occasional exceptions, but not very often.

It's a curve. Chances of failure go up in a curve over time. When the projected failure curve exceeds the projected revenue increase, then you've found the threshold. Sometimes adding 2 years of warranty is financially justified. Sometimes it's not.

I have 11.5 years of hard data right here in this \My Documents\Business\Infrastructure directory. Everything after my two year contract with IBM (1991-1993), and not counting my 1.5 year contract with Sun Microsystems in 2000-2001. It's what I use as an unbiased reference from every other company I've ever worked with (over 100 is all).

Those would be very close estimates based on thousands of deployments across several dozen large companies. I wouldn't have any reluctance in submitting a seven- or eight-figure budget request for server and network hardware to a CIO using those numbers as an accurate risk assessment.

There are other considerations infrastructure managers use as well, but the MTBF "reality adjustments" are among the first used in any bulk hard drive purchase consideration. The alternative is using base figures...1.2 million hours MTBF at 100% load = 137 years of expected reliability at 100% load. Every infrastructure and operations manager I've ever known call that utter bullshit. Which is why they use something much more relevant and meaningful - their own experience and that of their counterparts - over manufacturer ratings.

Reality in a server room with 200+ business-critical machines running 24/7 over ten years is much different than you could ever tell from manufacturer specs, or even third party data. The people who manage all the machines in those rooms are the bottom line experts for product reliability. They're the ones who use the products by the hundred, and they use more of them in total than anyone else in the world.

It's exactly why such factors (and many others) are used to calculate functional viability of very large corporate purchases. It's also why I use my corporate experience - and that of the operations managers in my cell phone address book - to make home PC buying decisions. The numbers become large enough to provide statistical relevance, rather than a single home user who buys one or two things from a particular manufacturer each year.

Instead of norton ghost, why not just set up a raid array?

because I'm too damned computer illiterate to know what a raid array is...

I'm not bad at throwing parts together or using basic software, but otherwise I'm not all that educated on computer programming.

Worst part is, despite my limitations, I am the best at computer knowledge compared to just about everyone I know, and so I'm stuck being the go-to person for problems...

you'll have to teach me - I'm always willing to learn

hey, you at least do backup which is a lot more than most people can say.

A raid array basically is two hard drives connected together so that one mirrors the other. That way if one dies, you replace it without any data loss. Of course, it's a bit more complicated than that (there's different schemes, etc) but it's a great solution for maintaining the integrity of your data. The only drawback that I see with it is you may be in trouble should there be a physical disaster since the origanl and backup are in the same physical case.

i've never had any problems with a hard drive before. until i got a couple of Maxtor hard drives. one of them has some bad sectors in the middle of the drive but still usable. the other just kept crashing and lost all data every time. other than that, i've never had problems with hard drives.

So if I understand correctly, using this setup, it perfectly protects you from hardware failure, but does not protect you from software failure. In this case, antivirus and "system restore" are your last hope should something go wrong on the software end...

I think I know what you're talking about now - I think it is a major selling point on new systems these days, since some people just don't know what else to do with a pair of 160+ gig hard drives that the salesguy is trying to push.

Sounds a lot easier than ghosting discs, since this raid array just works on its own after being set up...

Right now I just back files up via copy & paste to another internal drive when they're modified, and I ghost once every month or so, or after a major install.

The ghosting goes to portable hard drives which comes in handy if a disaster happens.

Well, what does your data say? I mean, are we talking actual reliability percentages, or what?

I really don't think those MTBF figures are inaccurate. Mostly they're misunderstood. No hard drive manufacturer is going to say that it's going to be 137 years on average before a drive fails. No drive is going to still be working after 100 years. They know that. That's not at all what they're claiming.

What they are claiming is over the expected lifetime of the drive (Seagate's expected lifetime seems to be 5 years), they MTBF per drive is a million hours. I'm not a statistics expert, but 10 drives are going to have a lot less (I would say 1/10th...).

Norton Ghost is a backup technology and RAID is not.

Ghost is for platform imaging - essentially a client system backup application. RAID provides hardware fault tolerance, and should NOT be relied on to forego backing up your system. If you have RAID, you'll still be backing up your data to external media.

It says what I've been posting. Anything company-specific is a matter of NDA, you know that. All the details provided here are more than what's necessary for a buying decision.

And if there's any doubt, anyone can do what I've done to get this information. Buy 10,000 hard drives (of all brands, models and interfaces) and run them continually for over 10 years in a one million user environment. Then analyze the results.

;)

A drive rated with "1.2 million hours MTBF" means the drive is projected by the manufacturer to encounter only one failure in 1.2 million hours of operation. This is vastly bogus. No hard drive is going to provide even half of that running duration without having - not only a discreet failure - but a complete and permanent one.

Hence the "reality adjustment". :)

"This Seagate SCSI drive will encounter only one failure in 12 years of operation."? Now that's much more realistic. This quote won't get you laughed out of a conference room.

(Also, I worked a 12-month contract in a building with a StorageTek sales and service office, and we talked about MTBF several times. They opened with laughter, and then said the same things I've repeated here.)

And HD manufacturers aren't hurting in any way from us having these kind of "reality adjustments". IT people do this a lot, with almost every technology we touch. Every industry needs somebody with their feet on the ground.

;)

Yea, scor. That was my main point that it still doesn't protect you from any type of physical disaster or the drive being lost or stolen. However, we are discussing hard drive failure in this thread. :p

True. Both technologies are relevant because both are associated with HDD data integrity - one to help prevent data loss, and one to restore data if it is lost.

That's not what they mean by "MTBF".

They calculate MTBF based on the estimated drive lifetime (5 years, in Seagate's case, I believe), or possibly only over its first year (although failures are actually higher the first year, in general, so you get a higher number if you calculate based on 5 or 10 years).

That is, over the first 5 years of a drive's life (assumed to be the drive's lifetime), how often is a drive going to fail, on average?

Every 114 drive years (1 million hour MTBF). This is a useful measurement. So with 100 drives running all the time, every 1.14 years, you'll have a drive failure, on average.

Again, drive manufacturers do not expect any drive to still be working after 100 years, and that's not what they mean with their MTBF measurements, because no one cares if a drive is going to be working in 100 years (in 100 years, finding a working drive controller that would work with IDE would likely be a trick...)

Yeah, we tried to "reinterpret" MTBF many ways over the years. But it all comes down to the fact that one MTBF rating is given for one product, not for 100 products.

The data they "use" can come from any combination of lab test results, field measurements or prediction models. But it doesn't change the fact that expecting anything close to 137 years to pass between discreet failures for one drive - and that's indeed what MTBF infers: mean time between failures for the drive its rating is applied to - is outrageously inaccurate.

But all's good...IT budget managers know all this and derive their own much more accurate fault tolerance models before they spend their money.

It's not a "reinterpretation", it's what the drive manufacturors mean by MTBF. It's the accurate interpretation.

I'm sorry to be "mean", but since you're a very smart guy (and so this probably doesn't happen much), I figure it won't bruise your ego too much: You simply don't understand what MTBF means.

A good explaination can be found from an old IBM page which talks about MTBFs. See question #4, in particular.

No worries. :)

We've looked at that for years, hoping to preserve any legimate "intent" drive manufacturers may have had with the rating.

There's nothing mysterious or complex about it. And reinterpretation isn't going to make the rating more relevant. It's well understood by all of us who make mass HDD purchases in the IT industry. One rating for one drive, and the ratings are vastly inaccurate.

Don't think that this is my position alone. My references are based on thousands of hard drive buying decisions, interviewing dozens of HDD manufacturers for clarifications and compiling the knowledge of hundreds of infrastructure managers in the IT industry.

As all of us have, you'll learn more the farther into professional IT you get.

And if you apply for a job in any IT operations division in which you'd be making hardware buying decisions, don't say out loud that you put faith in raw MTBF ratings. Say instead that you are able to calculate "adjusted MTBF ratings in order to facilitate smart buying decisions" of magnetic storage media.

Ok, well, I read each of these posts when they pop up on this thread, and I ask myself,

"Where the hell is a rating that an average joe computer upgrader can understand?"

Thread topic = "are these any good?"

Really, this is what started the thread, right?

I think that all of you can agree that the average joe doesn't really benefit from this MTBF number (a.k.a. Most Total Bullshit Factor number)

Why can't the companies have an accurate yet understandable rating system that applies to the life of just one drive (not many bought at once or several bought at 5-year intervals)?

The original post person was asking about the value of one drive, not 100 or 1000, and not the combined life/reliability of drives bought at 5-yr. intervals.

I have yet to find a consumer reports article on the issue of drive life and reliability, and usually I can trust them to give me a decent general unbiased rating for just about anything.

Maybe one of you readers has a link or access to some kind of unbiased report on drive life/reliability. If so, please post, so that we can all benefit from it.

As far as personal experience goes, I've probably owned around 30 drives, mostly around 20-80 gig, but a few up to 250 gig on the newer systems and portable storage. I have yet to have hardware failure, so I really don't know what brand is better than another. Where can I find out what's best?

It's there. I'm not saying it's accurate, but it's there.

There is some complexity to it, and the thing is that most drives will not fail over, say, 5 years.

I'm not saying that the ratings are accurate, by the way — I have no way of verifying that they are, but I think they most likely have some usefulness, and a fairly good accuracy.

And note that you need to know the manufacturer's drive lifetimes (among other things, but primarily that) to make useful comparisons of MTBFs...

Well, your understanding of it is frankly wrong; so perhaps the industry has a vast misunderstanding of MTBF. It wouldn't surprise me, the information is scarce, and I've even seen manufacturers websites get it wrong.

Actual idiotic quote:
That is not what a MTBF of 500,000 hours means. Not at all. It's not even correct if you fail to account for drive lifetimes. Any competent engineer working for Samsung knows this (any person with experience in the IT industry knows this); I predict one would find a PHB managing the technical content on the website.

I'm sure I would... but I hope to be hiring people for that sort of stuff (I'd rather be paying attention to The Code). :D ;)

I'll keep that in mind! :D :D

There isn't one.

Not really, no.

Because that would require them to run the drives for 5 years before releasing them to market, actually. Plus, that would open them up to frivolous lawsuits, and no hard drive manufacturer wants to say that their drive is significantly less reliable than their competitor...

Well, actually, MTBF is a fairly accurate measure of that (presuming MTBFs are actually accurate, which, again, I can't say if they are or not), you just need to understand both what it means and the limitations of statistics in predicting a single drive's life.

With a MTBF of 1,000,000 hours (114 years), a single drive has something like a 4% chance of dying (I did not do the correct calculations for this, but 4% is about right.)

http://www.storagereview.com/ is the only place I've ever heard of.

Hard drives are pretty reliable, actually, if you replace them before their estimated lifetime ends.

There's no silver bullet. The best you can do is listen to people who've bought and worked with many hard drives from many manufacturers over many years. The higher those numbers (# of drives, # of manufacturers, # of years) - and if you have a baseline of trust for the person - the more you can trust their recommendations as being statistically relevant.

With adjustment factors, it is indeed. :)

With adjustment factors, they are indeed.

Quite true. The MTBFs do have to have linear consistency so that factors can be applied with reliable results for making smart buying decisions.

I respect you're position. But I disagree, and thousands of IT hardware experts share my position on this.

We can't abandon so much consistency and so much relevant experience by all these people in the industry over decades when adjusting MTBF ratings.

If we were wrong more than a tiny fraction of the time then I'd be more inclined to doubt the methods used. But the number of inaccurate projections we've made using adjustment factors have been very low considering the number of products purchased. Most inaccuracies I've observed are when novice employees make purchase recommendations based on unaltered MTBF ratings.

In other words, if you go into information technology you'll see what I mean. Give it a few years. ;)


The information within companies on this subject is vast. Each company has volumes of data on product reliability collected both internally and from their business partners over many years.

Software engineers don't hire anyone. Most are entry-level right out of college. They're peers of network admins and DBAs, who make recommendations to infrastructure managers or directors who own the budgets used to make the actual purchases.

Unless you're talking about a development manager (who may also be a coder himself/herself). But that position takes a few years of experience after college to reach. It's definitely a good one to aspire to, though.

:)

With adjustment factors, they are indeed.

You need a duration criteria. Without one, all hard drives have a 100% chance of dying. (Well, actually there's no relevant statistic without a duration criteria.) :p

The factor depends on the manufacturer and reliability record of the line, but if you use a factor of 0.05...
A raw MTBF of 1 million hours * 0.05 = a projected full-load duration between discreet failures of 50000 hours = 5.7 years.

As you said before, theres no way of being totally sure since there's not a way to be sure of the accuracy of any particular MTBF rating by a manufacturer, but if the factor is relevant for a particular manufacturer/product line, then using this drive failure estimate is very reasonable and supportable to a person making a buying decision.

I'm coalating the numbers from some 70 of my most recent client companies to put together an accurate list of adjustment factors for all popular HDD manufacturers. I'll post it when it's done.



edit:

@xzxzxx: Please don't think I mean you any insult in all of this. I'm just standing by my position and experience, that's all. You know I love you, man. :)