[Nuke61]

waldojim, on 17 May 2012 - 11:58 PM, said:

The second source, with the pretty chart you posted, is running 20 drive arrays. Ok - I am starting to understand the sources are meant for Enterprise techs now. Understand - that 3 drives actually reduces my failure rate considerably first and foremost. Then lets cite the more appropriate graph:
http://media2.servet...MTTDL-RAID5.png
With a 2% chance of failure over 10 years, I think I will sleep just fine at night.

Though after reading all this, it looks like my next array will consist of 4 drives, and raid 10. There is something to be said for a <1% chance of failure.

I might be missing another difference, but the only difference I see is that the first graph assumes a 5 year MTBF, while the one you cite assumes a 10 year MTBF. But the actual failure rates wasn't the point of posting it; it was simply to show that RAID isn't equivalent to a backup.

[waldojim]

Nuke61, on 18 May 2012 - 06:10 AM, said:

waldojim, on 17 May 2012 - 11:58 PM, said:

The second source, with the pretty chart you posted, is running 20 drive arrays. Ok - I am starting to understand the sources are meant for Enterprise techs now. Understand - that 3 drives actually reduces my failure rate considerably first and foremost. Then lets cite the more appropriate graph:
http://media2.servet...MTTDL-RAID5.png
With a 2% chance of failure over 10 years, I think I will sleep just fine at night.

Though after reading all this, it looks like my next array will consist of 4 drives, and raid 10. There is something to be said for a <1% chance of failure.

I might be missing another difference, but the only difference I see is that the first graph assumes a 5 year MTBF, while the one you cite assumes a 10 year MTBF. But the actual failure rates wasn't the point of posting it; it was simply to show that RAID isn't equivalent to a backup.

With a <1% chance failure over multiple drive replacements, it sounds far more feasble than a single external backup. It seems to me that the story points out how an external backup is no more a goo backup han raid.

[Evildave]

The chance is non-zero.

And individual drive sizes are creeping up.

$270 for a 4TB HDD? Just back it up!
http://www.google.co...=4tb+hard+drive

[waldojim]

Evildave, on 18 May 2012 - 11:06 AM, said:

The chance is non-zero.

And individual drive sizes are creeping up.

$270 for a 4TB HDD? Just back it up!
http://www.google.co...=4tb+hard+drive

Or don't buy a 4TB drive.

[Evildave]

Or buy a 3TB drive for $123, or a 2TB for $109, or a 1TB drive for $90. They're pretty cheap. You're right: 3TB seems to be the price point to beat.

Or don't back it up, if that's what you really, really want.

You could do it, it would be cheap and convenient and quick, and you don't want to.

If it helps, I can tell you NOT TO back it up, so you don't feel like your back is against the wall?

[waldojim]

Evildave, on 18 May 2012 - 11:49 AM, said:

Or buy a 3TB drive for $123, or a 2TB for $109, or a 1TB drive for $90. They're pretty cheap. You're right: 3TB seems to be the price point to beat.

Or don't back it up, if that's what you really, really want.

You could do it, it would be cheap and convenient and quick, and you don't want to.

If it helps, I can tell you NOT TO back it up, so you don't feel like your back is against the wall?

OR I can keep using what already works, and has a lower chance of failure than a cheap POS backup drive with crappy little powersupply.

[Evildave]

Those are bare drive prices. Get any kind of toaster dock or box you like. Toaster docks are inexpensive, and include abundantly adequate power supplies. As are boxes to keep the drive in. Even the original box+static bag they shipped it in.

Where your RAID can fail in any number of ways, having another copy, separate from that contraption is where you get the safety.

Is it possible that external drive could fail? Yes. It's equally possible that any drive in your RAID can fail, too. And if you don't have a hot drive on standby, you'll be rolling the dice, running a degraded RAID that is just WAITING TO DIE, while you wait for the replacement drive to show up.

So why not buy the 'hot' replacement drive and USE IT to back up at lest the most 'important' parts of your RAID.

Of course, you'd feel pretty damned stupid if you had a drive failure in the RAID, and then a second failure while you're rebuilding it on top of your ONLY OFFLINE BACKUP.

Which is an oft-cited failure mode, actually. Since the whole batch of disks you get for your RAID usually came from the same run of the factory, if one fails, another will probably be hot on its heels to fail next.

So buy a hot spare, AND a backup drive. Really. It's not rocket science or brain surgery.

[Nuke61]

waldojim, on 18 May 2012 - 09:01 AM, said:

With a <1% chance failure over multiple drive replacements, it sounds far more feasble than a single external backup. It seems to me that the story points out how an external backup is no more a goo backup han raid.

Is the rate of failure really <1%? That rate is based on using the 10 year MTBF rate for the drives, but it's far more likely that people are using consumer grade disks. The author pointed out that that the reason why he included it was to show the dramatic effect that MTBF has on predicted failure rates.

Additionally, as the article points out and Google's real world experience shows, the real world failure rate is far worse than the ideal numbers given by the manufacturers. As far as a single backup drive goes, I agree and think a RAID5 setup would be a better "backup" strategy than a single external drive. But then again, I don't rely on a single backup drive - I have 3 of them, one of which is a mirrored RAID NAS. I also use Time Machine and cloud backup, but I admit I'm bit of a backup nut

[waldojim]

Nuke61, on 18 May 2012 - 01:59 PM, said:

waldojim, on 18 May 2012 - 09:01 AM, said:

With a <1% chance failure over multiple drive replacements, it sounds far more feasble than a single external backup. It seems to me that the story points out how an external backup is no more a goo backup han raid.

Is the rate of failure really <1%? That rate is based on using the 10 year MTBF rate for the drives, but it's far more likely that people are using consumer grade disks. The author pointed out that that the reason why he included it was to show the dramatic effect that MTBF has on predicted failure rates.

Additionally, as the article points out and Google's real world experience shows, the real world failure rate is far worse than the ideal numbers given by the manufacturers. As far as a single backup drive goes, I agree and think a RAID5 setup would be a better "backup" strategy than a single external drive. But then again, I don't rely on a single backup drive - I have 3 of them, one of which is a mirrored RAID NAS. I also use Time Machine and cloud backup, but I admit I'm bit of a backup nut

I understand that, and for what it is worth, he did mention that 5 years was a more realistic expectation. As I have mentioned elsewhere, I have already established a 3 year replacement cycle. So sometime this year, all three drives will be replaced, and go into a new array with at least double the total space. Though, thanks to that article, I may be convinced to move either to raid 6 or 10. I kind of like the idea of having a 2 drive fault tolerance.

[Evildave]

So buy a $40 toaster dock, and 'retire' the disks to be backups. They'll still have 'a year or two' of expected life left on them, and divided by an hour or so of use a week, that's a lot of time left.

[waldojim]

Evildave, on 18 May 2012 - 02:49 PM, said:

So buy a $40 toaster dock, and 'retire' the disks to be backups. They'll still have 'a year or two' of expected life left on them, and divided by an hour or so of use a week, that's a lot of time left.

nope, they will go into other machines as "junk" drives. I trust a brand new set of drives going through a monthly data sanity check on a trusted raid controller far more than old drives on a junk power supply. Not sure why you have a problem with this, but get over it.

Also, as the array grows in size, the 2TB array I have now is not going to be sufficient for backups.

[Evildave]

These are thoroughly 'burned in' and still somewhere at the bottom of the 'bathtub curve'.

Brand new drives are absolutely THE MOST prone to failure.

[LiveBrianD]

waldojim, on 18 May 2012 - 04:55 PM, said:

Evildave, on 18 May 2012 - 02:49 PM, said:

So buy a $40 toaster dock, and 'retire' the disks to be backups. They'll still have 'a year or two' of expected life left on them, and divided by an hour or so of use a week, that's a lot of time left.

nope, they will go into other machines as "junk" drives. I trust a brand new set of drives going through a monthly data sanity check on a trusted raid controller far more than old drives on a junk power supply. Not sure why you have a problem with this, but get over it.

Also, as the array grows in size, the 2TB array I have now is not going to be sufficient for backups.

Unless you're buying the cheapest, most generic external enclosure you can possibly find, the power supply is NOT likely to be an issue. Second, since you have an array anyway, why replace the drives every 3 years, when they probably still have a lot of life left? Come to think of it, the WD Green 1TB in the tivo is from 2009 I think (yep, 3 years) and has been run 24/7, and it has no issues. Yep, a low-end consumer drive with almost no downtime, and it's doing just fine.

[waldojim]

LiveBrianD, on 18 May 2012 - 05:42 PM, said:

Unless you're buying the cheapest, most generic external enclosure you can possibly find, the power supply is NOT likely to be an issue. Second, since you have an array anyway, why replace the drives every 3 years, when they probably still have a lot of life left? Come to think of it, the WD Green 1TB in the tivo is from 2009 I think (yep, 3 years) and has been run 24/7, and it has no issues. Yep, a low-end consumer drive with almost no downtime, and it's doing just fine.

How many external enclosures have you used? I have owned everything from the cheap little "simpletech" drives, to custom high end enclosures. The ONLY solid boxes I have owned that didn't blow up and toast their drives (after 2 years)- were top quality SCSI boxes. Not the slightest bit useful in this situation. The simpletech box lasted 6 months.

EDIT: Those WD Green drives have roughly 5 year lifespan with increasing chance of failure after year 3. Be happy it still works, but don't be surprised when it dies. I will be just fine, and quite happy with 3 brand new 2TB drives when I switch. Until the, I will be happy with what I have. Namely because - it works, and has worked reliably for some time now.

This post has been edited by waldojim: 18 May 2012 - 05:54 PM

[Evildave]

Both my 'toaster dock' ones still work fine. Turns ANY stray hard disk into backup or transportation media.

Oh, and no problems with 'ventilation'. After all, the naked hard drive is right there exposed to the air.

This post has been edited by Evildave: 18 May 2012 - 05:57 PM

[waldojim]

Evildave, on 18 May 2012 - 05:56 PM, said:

Both my 'toaster dock' ones still work fine. Turns ANY stray hard disk into backup or transportation media.

Oh, and no problems with 'ventilation'. After all, the naked hard drive is right there exposed to the air.

Ventilation isn't a problem. It has been proven that hard drives are fine up to 50C.

[Nuke61]

Evildave, on 18 May 2012 - 04:58 PM, said:

These are thoroughly 'burned in' and still somewhere at the bottom of the 'bathtub curve'.

Brand new drives are absolutely THE MOST prone to failure.

According to Google's study of huge populations of hard drives, the 'infant mortality' is of marginal consequence. Per Google's report of their own hard drive population, the 3 month annualized failure rate (AFR) is 3%, and it drops to just under 2% at 1 year. From that point forward approximately 7% of the remaining drives from the initial population fail each year. In other words, new drives are slightly more prone to failure than a 1 year old drive, but they are not the most prone to failure. The older the drive, the more likely it is to fail, with the small exception of a brand new drive.

Everything You Know About Disks Is Wrong: http://storagemojo.c...disks-is-wrong/

Google study: http://static.google...sk_failures.pdf

[LiveBrianD]

waldojim, on 18 May 2012 - 05:52 PM, said:

LiveBrianD, on 18 May 2012 - 05:42 PM, said:

Unless you're buying the cheapest, most generic external enclosure you can possibly find, the power supply is NOT likely to be an issue. Second, since you have an array anyway, why replace the drives every 3 years, when they probably still have a lot of life left? Come to think of it, the WD Green 1TB in the tivo is from 2009 I think (yep, 3 years) and has been run 24/7, and it has no issues. Yep, a low-end consumer drive with almost no downtime, and it's doing just fine.

How many external enclosures have you used? I have owned everything from the cheap little "simpletech" drives, to custom high end enclosures. The ONLY solid boxes I have owned that didn't blow up and toast their drives (after 2 years)- were top quality SCSI boxes. Not the slightest bit useful in this situation. The simpletech box lasted 6 months.

EDIT: Those WD Green drives have roughly 5 year lifespan with increasing chance of failure after year 3. Be happy it still works, but don't be surprised when it dies. I will be just fine, and quite happy with 3 brand new 2TB drives when I switch. Until the, I will be happy with what I have. Namely because - it works, and has worked reliably for some time now.

Right now, there are 3 WD Elements drives in the house, and all work perfectly. Admittedly, I had a simpletech that failed - it didn't kill the drive or anything, it simply started corrupting file transfers and giving errors - the files on the drive were otherwise fine, and the drive is now working perfectly as an internal drive in a spare machine (fyi, the 12V power supply that came with it also works, and is currently being used to run a bunch of old computer fans I taped together on hot summer days). I also had a lacie NAS that failed (again, the drives were fine, one is in my PC right now and the other is in the closet because I don't need it), though that had a lot of reviews noting that the PCB is prone to failure in it. (one day, it failed to start and spin up the drives - it took me 20 tries to get it to start so I could copy a few files off it that my dad hadn't backed up) As long as you buy something with good reviews and all, it shouldn't be a problem. And as I said, all of the drives I had in enclosures still work, even if a few of the cheap enclosures don't. (fyi, both the simpletech and the lacie were several years old when they failed, though they weren't on all the time).

Fyi, the Samsung drives I have (which were in the lacie nas) are from mid-2007 - yep, 5 years (and smart says they're perfect, they're working just fine), and the WD is from 2005 I think (it's IDE, so that gives you some idea - SATA wasn't new when my dad bought that one though) - 7 years, and working fine.


Also, are you saying a 5 year lifespan when on 24/7 or, say, 10 hours and a few cycles per day? (which is more typical in a PC)

[davebav50]

Always amuses me when someone says "it wont scale" or there is an end date. I am getting on a bit now and remember doing Transistor theory at college years ago, the lecturer was adamant that todays microchip performance couldnt happen because the physics of the chip and the wavelengths of the high frequencies involved would prevent it. Similar was said about the data limits of telephone cables, not even 1mb/s would be possible! I used to work on high capacity data circuits that were a mere 9600! Magnetic hard drives were miniscule by todays standards and there were limits that couldnt be overcome - these were less than a few mb!
No, solid state drives will be here to stay, its just no one can foresee the ingenuity this week that is needed.