What hard drives to choose ?

[cyberjock]

@ Noobsauce,
We will have to agree to disagree. I know the Samsung 2TB drives I had been running has a mode for parking the heads while leaving the drive running, it's very obvious when you hear the heads loading. This is not Intelli-park and as far as I know the drive must be put into this lower power mode. And as for the wear leveling, I've read about that in the past quite a bit and I've never heard of it on a non-A/V drive but I won't state it doesn't exist on a non-A/V drive, and everything I've read. Here is an example:
http://www.wdc.com/wdproducts/library/AAG/ENG/2178-771127.pdf

Which states: Preemptive Wear Leveling (PWL)
The drive arm frequently sweeps across
the disk to reduce uneven wear on the
drive surface common to audio video
streaming applications.

I've never heard of the arm being moved for thermal stabilization benefits and I don't recall the heads on my computer drive just sweeping across the platters without some sort of cause. And the heads do crash periodically, it happens and with these mechanical drives there is no way around it. On some drives, like the newer laptop drives, a sudden impact forces them to park the heads immediately. The heads pop back out quickly after that event but it's an attempt to reduce the damage inflicted by the user.

I do know back in the 1999-2002 I had some drives(I forget which brand exactly) would park the heads after a certain amount of idle time in an attempt to increase reliability. The theory was that if you park the head for much of the drive's life it might extend their lifespan by protecting the data from potential damage from the head. I haven't had any drives do that since around that time, and I had been told years ago it was because it was determined there was observed benefit.

Interesting that they'd even make a comment like that(about PWL). I'd really like to know how a single sweep of a mechanical arm will "even" out the wear and tear. In AV you usually have a constant even reading or writing, and based your setup you could easily use some sectors more than others. Honestly, it sound ALOT like marketing BS. I'll have to send an email to my contact and see what the hooplah is about.

To be honest, I'm not sure believe that the Intelli-park technology increases reliability either. But I have no data to back up that claim, just my opinion.

As for the temperature compensation, the data physically moves as the platters and mechanical arm expand. The temperature is taken and a few disk reads are performed to determine how much compensation is required. I think that the industry standard was something like 1 compensation every 5 minutes for the first hour. When the timer determines that the compensation is required it is not normally performed until the next read or write is performed, but some drives will do it after the timer is up regardless(it's mostly based on manufacturer). For the drives that wait, the compensation takes 20-30ms, so you would likely never be able to notice that the compensation is being performed.

 

[Cooledspirit]

@ Noobsauce,
We will have to agree to disagree. I know the Samsung 2TB drives I had been running has a mode for parking the heads while leaving the drive running, it's very obvious when you hear the heads loading. This is not Intelli-park and as far as I know the drive must be put into this lower power mode. And as for the wear leveling, I've read about that in the past quite a bit and I've never heard of it on a non-A/V drive but I won't state it doesn't exist on a non-A/V drive, and everything I've read. Here is an example:
http://www.wdc.com/wdproducts/library/AAG/ENG/2178-771127.pdf

Which states: Preemptive Wear Leveling (PWL)
The drive arm frequently sweeps across
the disk to reduce uneven wear on the
drive surface common to audio video
streaming applications.

I've never heard of the arm being moved for thermal stabilization benefits and I don't recall the heads on my computer drive just sweeping across the platters without some sort of cause. And the heads do crash periodically, it happens and with these mechanical drives there is no way around it. On some drives, like the newer laptop drives, a sudden impact forces them to park the heads immediately. The heads pop back out quickly after that event but it's an attempt to reduce the damage inflicted by the user.

@David,
I'm glad you are diving into the details of your parts before you purchase them but if you are going to build a NAS from scratch and buy new drives, I strongly suggest you buy the WD Red drive lineup. They are new but they are also a NAS certified drive, meaning they can handle 24/7 abuse. They are also low power and very quiet. The cost is the same as a good quality desktop drive as well.

I have just tested the drives using ataidle command "atailde -S 30 /dev/ada0" to verify the drives can be commanded to sleep, used the "smartctl -A /dev/ada0" to check the spinup count before and after and it increments. I also used the -n standby switch and it recognized the drive was in standby. I have setup my FreeNAS settings to sleep after 30 minutes and when I get home tonight I'll know if they went to sleep on their own. Current spinup count is 17 for my ada0 drive. If it changes then the drives must have spun down.

I'm not sure what else you need to know.

@JoeSchmuck: thanks so much for your input. I now know everything I wanted to know !

 

[cyberjock]

I heard back from my buddy about PWL. He said that its been turned into a marketing gimmick after it was realized that there really is no benefit for real world performance. The theory was that any particles on the platter could be "brushed" off by slowly sliding the head over the surface of the disk. Unfortunately most particles seem to be either too small to be brushed off or so large that attempting to brush them off can damage the media itself. Additionally most particles seem to be commonly brushed off by the high rotational speed of the platters.

He said that the firmware change to add PWL wasn't difficult and although it would be easy to remove, marketing hype is what is keeping the "feature" around.

So yeah.. I'm not sure how to take it.

 

[joeschmuck]

Now I have no problems believing it's a marketing gimmick thing at all because I've nly heard of this for A/V drives. Why not have it on desktop drives as well just to all a bit more longevity or NAS drives that get minimal use?

So I believe you.

Oh, my WD Red drives do spin down after the requested time limit through FreeNAS. The only problem I see is if someone is using an active jail then the drives are likely to either remain spinning or spin up and down too frequently. My jails are on a USB Flash Drive for this reason alone.

 

[jgreco]

I'd really like to know how a single sweep of a mechanical arm will "even" out the wear and tear. In AV you usually have a constant even reading or writing, and based your setup you could easily use some sectors more than others. Honestly, it sound ALOT like marketing BS.

Without commenting on the relative merits of various technical and marketing techniques for making drives more valuable, I will note in passing that in my office, I do have an 8" hard drive sitting on display, cracked open, and you can clearly see the wear pattern towards the middle of the platters. I am going to make a bit of a wild guess here and suggest that sending the heads out to idle over random tracks might be a reasonable technique to help reduce wear. How to do that in a practical manner is left as an exercise for the reader.

 

[cyberjock]

I used to have a drive like that years ago. It was bulky and heavy. We did, after about an hour, get ours apart to look inside. It had marks on the inner part of the platters. We assumed it was the landing zone. My friend said that the material is radically different from what it was 30+ years ago and that looking at old media from that time period is not a good indicator of problems the current generation have.

He said if you really want to see what a worn out disk looks like, open up a failed disk in the last 3-4 years. The perpendicular recording required some material changes, and the type of wear you and I saw would have destroyed a current generation. Wear products have to be extremely small because of the tolerances, so the disk is designed so that wear products shouldn't normally build up inside the hard drive. I will admit I opened up a disk about a year ago that was failed. There was no indication of any wear inside the drive. But I also couldn't vouch for how used the drive was. I just knew it was out of warranty and wouldn't work.

 

[joeschmuck]

I opened up a drive about 2 years back, was running fine but also it wasn't new. I cut a sizable window into it and attached a Lexan sheet with some silicone. It still runs fine and I'll pull it out to show someone how fast the heads move around. Nope, no clean room and I don't expect to put my data on it ever. I'd hope the filter would screen out any contaminants I added. So far it's still a mirror shine on the top platter that I can see.

Yes, materials have changed substantially in the past 30 years which includes substrates as well.

 

[jgreco]

While both those points are valid, there's also some validity to the idea that continuously running the heads over a single track is more likely to induce some form of wear than if the load is spread out; whether or not this actually matters for longevity is of course another matter, since if it only becomes a factor after, say, 20 years, then that's beyond the effective service life of the drive.

It's always great to say "don't look at yesterday's tech, today's doesn't have that problem", because sometimes it's true, and more frequently it's a complicated lie, because in almost all cases, today's tech will have new and more interesting failure modes, often in addition to many of the old ones. Realistically, to see if wear patterns become visible in today's drives, we could plug one in and let it run for ten years, and then resume this discussion in ten years. That's not practical. So from an engineering perspective, I'd rather not bother, and instead, note that it's safe to assume it could be a problem, one which could be painlessly remediated by randomly choosing a track over which to idle. On the other hand, we used to assume that drives would be used for many years, while today a three-year lifecycle is often considered the "norm". So... lots of things to consider, and not really any obvious conclusions. Except that artificial wear won't help longevity.

 

[joeschmuck]

@Cooledspirit

Did we answer your question to your satisfaction. I know we got a bit off topic here but I hope you now have a better handle on what drive to purchase for your system. Also, you could just go buy some enterprise high dollar drives if you have a lot of money

 

[Cooledspirit]

@Cooledspirit

Did we answer your question to your satisfaction. I know we got a bit off topic here but I hope you now have a better handle on what drive to purchase for your system. Also, you could just go buy some enterprise high dollar drives if you have a lot of money

Thanks both... I know enough for the time being :)

Currently I'm trying to find out if my future hardware (non-hdd :p ) is supported.

Regards,


Cooledspirit.