Hello,

 

I've been reading up on Power on Hrs for HDD's. I now have 7xHDD's in my server that have done over 20000 hours. Under the SMART page from unMenu there is some blurb about HDD's rated at 5 years (approx 43000), but other searches etc have turned up little info?

 

Should I be concerned? Thx

No, you shouldn't be concerned.

 

I have several drives with well over 50,000 hours that are still just fine.

 

A drive accruing power-on hours is kind of like a human who keeps getting older.    Not a lot you can do about it 

 

At 20,000 hours, the drives are not even middle aged. 27,000 is ~3 years.

 

Gary's 50,000 hour drives are less likely to wake up again after a power cycle. But that does not mean they are likely to fail if left running in a stable environment.

 

Drive age has not been shown to be an indicator of failure.

 

As an interesting side note, at least one drive manufacturer does have drives that self disable when power cycled if projected life is exceeded. That can make for a very bad day in array management.

Thx very much for the insight. I've been a little worried about the 20000+ ones, so it's good to know. One of the reasons I'm updating my parity drive (see other thread) is that I trying to reduced overall HDD numbers from 16 drives to eight. Hopefully, that'll reduce the time it takes to do a parity check (or HDD rebuild). HDD's are so large now that I can double my existing storage with only 8 drives.

 

Wow...50000 hours seems a lot 

 

Thx again

Hopefully, that'll reduce the time it takes to do a parity check (or HDD rebuild).

 

Actually it will make these times LONGER.    If the drives have equivalent areal density it would be twice as long; although it's likely your older drives have lower areal densities, so it won't be quite that bad.    But unless the larger drives double your areal density, then the parity checks and rebuilds will be longer -- not shorter.

 

Hopefully, that'll reduce the time it takes to do a parity check (or HDD rebuild).

 

Actually it will make these times LONGER.    If the drives have equivalent areal density it would be twice as long; although it's likely your older drives have lower areal densities, so it won't be quite that bad.    But unless the larger drives double your areal density, then the parity checks and rebuilds will be longer -- not shorter.

 

Hey Gary, Thx for the input, but I still don't understand this. Are you saying that even if I reduce the number of drives, parity checks could still take as long?? I thought the PCI bottleneck (for the TX4 cards) was one of the main reasons the check took so long.

 

When I upgrade from a 2Tb parity to a 4tb parity the check jumped from approx 22hours to 27+ hours. Both are still to long

Reducing the number of drives might help reduce the parity check, if you are removing the slow drives. If you remove some drives, but leave the slowest, the parity check will not change much.

Okay, you've got a significant bottleneck in the bandwidth to your drives, so that also comes into play here.    Makes the math a bit more complex with regard to what will happen as you decrease your drive count.

 

Looking at your configuration, I see the following:  7 SATA ports on the motherboard, 8 more ports on the 2 TX4 cards.  But you indicate you have 16 drives.  What am I missing?  [Do you have a single PATA drive?]

 

There are several things that come into play with parity check speeds.

 

First, if any of the interfaces can't sustain the data rates that the drives can sustain (NOT the interface speed -- the sustained data rate), then the interface will cause things to slow down.    That's almost certainly the case with your TX4's, assuming you have 4 drives connected to each of them.    If you reduce the drive count on the TX4's to 2 drives on each, that will significantly improve your speeds.

 

Next, the areal density of the drives you're using.    If you're using old drives with only 160GB/platter areal density, they clearly aren't going to have sustained data rates that come anywhere close to modern 1TB/platter drives.    Common platter sizes in recent years are 160GB, 250GB, 320GB, 500GB, 667/750/800 GB, up to 1TB ... with the newest 6GB WD Reds using 1.2TB platters.    The higher the density, the higher the sustained data rate, given the same rpm rate.

 

The rotational speed of the drives also comes into play, but actually has far less of an impact than the areal density.

 

Finally, the SIZE of the drives matters -- or, more accurately, whether or not you're mixing sizes in your array.  Every drive has much faster sustained data rates on the outer cylinders, since a lot more data/rotation is available to "feed" the processor.  As the drive moves towards the inner cylinders, the sustained data rate decreases markedly.    When you do a parity check (or drive rebuild), the speed can never be faster than the slowest speed of any drive currently involved in the check.  So if you have mixed drive sizes, as each drive nears the inner cylinders, the check will slow down a good bit until you pass the size of the drive that was limiting the speed.    If all of your drives are the same size, you only encounter this issue once;  but if you have a mix of drive sizes, you could encounter this several times.

 

 

So ... in your specific case ==>  The first thing I'd try to do is remove as many drives as you can from the TX4's.  One drive on each would definitely not be an issue;  and even 2 drives would not likely have much impact [2 modern 1TB/platter drives would still be bandwidth limited on the outer cylinders, but not by a lot].

 

If you get down to 9 drives, with 7 on motherboard ports and one on each TX4, you won't have any bandwidth limitations, so the parity check/rebuild speeds will be purely based on the sustained data rates of the drives.    In this case, you simply want to be sure you're using drives with 1TB/platter (or higher) areal densities.    I see from your other thread you're already discussing using a WD Red 6TB unit for parity.    This is a good choice, and has excellent areal density (1.2TB/platter) ... but if your goal is the fastest possible parity checks, AND if 4TB drives would provide you with as much space as you want [with 9 drives 4TB units would give you 32GB of protected space, or 40GB if you used 11 drives with 2 on each TX4],  then you'd get faster parity checks with 4TB drives, since there's only 2/3rds as many bits for the parity check to traverse.  EITHER of these should give you parity checks much faster than you're seeing now, however.

 

Your #1 goal should be to (a) get rid of any really small, low areal density drives you're using (especially if you still have a PATA drive in the mix);  and (b)  get the drive count on the TX4's down to no more than 2 drives/card (and 1 is even better).

 

A couple other thoughts ...

 

=>  If you list the exact make/model of all of your drives, I can tell you which ones you should focus on getting out of the mix first.

 

=>  Have you already bought a 6TB WD Red for parity?  [if so, that's fine ... but if that's the case, you may want to just buy 6TB data drives as well]

 

=>  How much total storage do you currently have (the list of drives will implicitly tell me that, of course) ... and what size are you hoping to build up to?

 

=>  Do you have any plans to upgrade the CPU/motherboard you're using in the relatively near future?  [Not necessary to do this -- the key benefit would be PCIe slots that might let you use a much better SATA controller if you need more ports than the motherboard has]

 

 

 

Think about Taxi cabs in NYC. They typically run many many more miles than a typical car.

 

Drives are somewhat the same. If you use them a lot, they tend to last longer and go further than if you are using them sporadically.

 

I personally think the wall clock age as opposed to the power on hours is more telling. At about 5 years there is a sharp increase in failures of some drives. But at about 5 years the drive sizes have increased so much (at least for the past 10-20 years) that a 5 year old drive is not such a valuable commodity and may be time to be retired anyway. We did get stuck at 3-4T for longer than usual, but are now seeing large leaps in size increases to 6T, 8T and soon 10T and 12T. Those 2T and 3T have or soon will be turning 5 soon and heading out to pasture.

 

Looking at your configuration, I see the following:  7 SATA ports on the motherboard, 8 more ports on the 2 TX4 cards.  But you indicate you have 16 drives.  What am I missing?  [Do you have a single PATA drive?]

 

Hey Gary, Really appreciate the expanded answer. It's helped a great deal.

 

The short answer here is no PATA Drive. My Mobo has an eSATA that is accessed from the back panel (making it 8xonboard ports).

 

My plan is to remove the 2xTX4 cards completely and leave 1xparity and 7xdata drives in the system. I currently run 21Tb, so using 7x4tb drives would only give 28tb (an amount of space in the short term that could be filled quite easily.)

 

Yes...I have purchased a 6Tb WD Red for the new parity, but as in my other thread, this drive is NOT being recognised in my system. I have carried out an RMA on it, but there is always the possibility that they find it okay - so not sure what I'll do then.

 

My drives are assigned as follows. The WD7500AAKS are obviously the oldest.

 

parity WDC_WD40EFRX

disk1  WDC_WD20EARS

disk2  WDC_WD20EARX

disk3  WDC_WD20EARS

disk4  WDC_WD20EARX

disk5  WDC_WD20EARS

disk6  WDC_WD20EARS

disk7  WDC_WD10EADS

disk8  WDC_WD7500AAKS (On Promise TX4 Expansion Card)

disk9  WDC_WD7500AAKS (On Promise TX4 Expansion Card)

disk10      WDC_WD7500AAKS (On Promise TX4 Expansion Card)

disk11      WDC_WD7500AAKS (On Promise TX4 Expansion Card)

disk12      WDC_WD10EACS    (On Promise TX4 Expansion Card)

disk13      WDC_WD10EACS    (On Promise TX4 Expansion Card)

disk14      WDC_WD10EACS    (On Promise TX4 Expansion Card)

disk15      WDC_WD20EZRX    (On Promise TX4 Expansion Card)

 

 

Think about Taxi cabs in NYC. They typically run many many more miles than a typical car.

 

Drives are somewhat the same. If you use them a lot, they tend to last longer and go further than if you are using them sporadically.

 

I personally think the wall clock age as opposed to the power on hours is more telling. At about 5 years there is a sharp increase in failures of some drives. But at about 5 years the drive sizes have increased so much (at least for the past 10-20 years) that a 5 year old drive is not such a valuable commodity and may be time to be retired anyway. We did get stuck at 3-4T for longer than usual, but are now seeing large leaps in size increases to 6T, 8T and soon 10T and 12T. Those 2T and 3T have or soon will be turning 5 soon and heading out to pasture.

 

Not sure I'll need anything more than 6Tb in the short term. 10Tb...wow!!!

 

My WD7500AAKS are the oldest in the system (they replaced some 500Gb Seagates). I built the box in 2007, so they are probably quite a few years old now. I don't run my server 24/7.

You have a couple of significant bottlenecks in your current configuration ...

 

#1 =>  As I noted earlier, having 4 drives on the TX4's is a significant bandwidth throttle

 

#2 =>  A LOT (most) of your drives have very low (by current standards) areal densities.

 

-- Your WD7500AKS drives only have 250GB platters ... that's 1/4th the areal density (and thus 1/4th the sustained data rate) of your 1TB/platter parity drive -- so even if it was on a motherboard port it would slow down parity checks and rebuilds a LOT. 

 

-- Your WD10EACS units may also be 250GB platters => that model was manufactured with both 250GB platters and with 333GB platters, so exactly which you have depends on the 6-character detail that follows the model number.  But either way, it's still much slower than current drives.

 

-- Your WD10EADS drive could have 333GB, 500GB, or 667GB platters, depending on when in the production cycle for this drive it was made (Again, the 6-characters after the model # would identify that).    Most of these were 2 platter drives with either 500GB platters or 667GB platters that only used the outer 2/3rd of the platter (which made them faster).

 

--  Your WD20EARS and WD20EARX drives are either 500GB platters (with 4 platters) or 667GB platters (with 3 platters)

 

--  The WD20EZRX is most likely a 1TB/platter drive, although there were apparently some early versions with 3 667GB platters

 

--  Your WD 40EFRX is a 1TB/platter drive

 

So the ONLY drives you have with an areal density of 1TB/platter are the 4TB WD Red (your parity drive) and probably your WD20EZRX.

 

Clearly you can make a BIG improvement in your array's speed by both moving off of the TX4's and moving to 1TB/platter drives.    You don't HAVE to use the largest drives ... but clearly that's the best way to reduce your drive count.  [Note that ALL WD Reds have at least 1TB/platter area densities => the 1TB, 2TB, 3TB, and 4TB units use 1TB platters; the 6TB uses 1.2TB platters]    Since you've already bought a 6TB Red for parity, I'd buy 3 more of those and change your array to a 6TB parity with 3 6TB data drives plus your current 4TB Red.    This would give you 22TB of space ... a bit more than you have now ... and you could add additional 6TB drives as your needs grow (since the price is likely to be lower over time, the additional drives would likely cost less).

 

... and just use all your current drives for backups of your data 

 

You have a couple of significant bottlenecks in your current configuration ...

 

#1 =>  As I noted earlier, having 4 drives on the TX4's is a significant bandwidth throttle

 

#2 =>  A LOT (most) of your drives have very low (by current standards) areal densities.

 

-- Your WD7500AKS drives only have 250GB platters ... that's 1/4th the areal density (and thus 1/4th the sustained data rate) of your 1TB/platter parity drive -- so even if it was on a motherboard port it would slow down parity checks and rebuilds a LOT. 

 

-- Your WD10EACS units may also be 250GB platters => that model was manufactured with both 250GB platters and with 333GB platters, so exactly which you have depends on the 6-character detail that follows the model number.  But either way, it's still much slower than current drives.

 

-- Your WD10EADS drive could have 333GB, 500GB, or 667GB platters, depending on when in the production cycle for this drive it was made (Again, the 6-characters after the model # would identify that).    Most of these were 2 platter drives with either 500GB platters or 667GB platters that only used the outer 2/3rd of the platter (which made them faster).

 

--  Your WD20EARS and WD20EARX drives are either 500GB platters (with 4 platters) or 667GB platters (with 3 platters)

 

--  The WD20EZRX is most likely a 1TB/platter drive, although there were apparently some early versions with 3 667GB platters

 

--  Your WD 40EFRX is a 1TB/platter drive

 

So the ONLY drives you have with an areal density of 1TB/platter are the 4TB WD Red (your parity drive) and probably your WD20EZRX.

 

Clearly you can make a BIG improvement in your array's speed by both moving off of the TX4's and moving to 1TB/platter drives.    You don't HAVE to use the largest drives ... but clearly that's the best way to reduce your drive count.  [Note that ALL WD Reds have at least 1TB/platter area densities => the 1TB, 2TB, 3TB, and 4TB units use 1TB platters; the 6TB uses 1.2TB platters]    Since you've already bought a 6TB Red for parity, I'd buy 3 more of those and change your array to a 6TB parity with 3 6TB data drives plus your current 4TB Red.    This would give you 22TB of space ... a bit more than you have now ... and you could add additional 6TB drives as your needs grow (since the price is likely to be lower over time, the additional drives would likely cost less).

 

... and just use all your current drives for backups of your data 

 

Thx Gary for the really insightful post. When you never grew up with computers, some of this stuff is way above my head. Your post here exactly equals my thinking. Use a new 6Tb parity, then replace disk7 (1tb) with the old 4Tb parity. That'll take care of all the 750Gb drives (once I copy all the data across). Later, a couple more 6Tb drives will take care of extra space and the remainder of the other drives on the TX4's. Hopefully, the ultimate aim is reduce parity / rebuild times, and to reduce the statistic of a drive failure by having less drives.

 

Very much appreciated. Thank you.

... Use a new 6Tb parity, then replace disk7 (1tb) with the old 4Tb parity. That'll take care of all the 750Gb drives (once I copy all the data across) ...

 

That's a good first step.  One more thing:  After you've done that, move two of the drives from the 2nd TX4 to the first one, so you only have two drives on each of the TX4's.    That will largely mitigate the bandwidth restriction your 2nd TX4 is causing on the drives attached to it.

 

... then when you later add another 6TB drive (to replace one of your 2TB EARS units) you'll be able to copy the data from the 3 1TB drives on the TX4's to it, and can then remove those 3 drives and one of the TX4's => at that point you'll only have one drive attached to a TX4 ... which is fine.    At that point your only performance limitation will be the areal density of the remaining 2TB drives ... but your performance will be MUCH better than you're seeing now, so you can likely just live with it and replace the remaining drives at a more leisurely pace -- just buying new 6TB drives as you need more space.

 

... Use a new 6Tb parity, then replace disk7 (1tb) with the old 4Tb parity. That'll take care of all the 750Gb drives (once I copy all the data across) ...

 

That's a good first step.  One more thing:  After you've done that, move two of the drives from the 2nd TX4 to the first one, so you only have two drives on each of the TX4's.    That will largely mitigate the bandwidth restriction your 2nd TX4 is causing on the drives attached to it.

 

... then when you later add another 6TB drive (to replace one of your 2TB EARS units) you'll be able to copy the data from the 3 1TB drives on the TX4's to it, and can then remove those 3 drives and one of the TX4's => at that point you'll only have one drive attached to a TX4 ... which is fine.    At that point your only performance limitation will be the areal density of the remaining 2TB drives ... but your performance will be MUCH better than you're seeing now, so you can likely just live with it and replace the remaining drives at a more leisurely pace -- just buying new 6TB drives as you need more space.

 

Thx Gary, I'll do just this.

 

One question. Unraid keeps track of the drives via serial number and not the port designation (is that right)?? So that means I can change the ports without issue.

That's correct => you can freely move the drives to any port you want.

 

Wow, one of my ST2000DL003 is at 75,000.