HDD Advanced Power Management

[sonny81]

I'm looking to upgrade my system but I want to be sure I'm not wasting money (please bare with me).

I understand that everyone has different experiences with many hard drive manufacturers so I hope for an unbiased response.

For me, Seagate Barracudas have been the only drives I've owned in a 9year period that have never failed. That said, I tend to need larger drives and replace old ones every 3-4years.

Anyway, I read an interesting article about Freenas being a "disaster waiting to happen when using desktop drives." Reality is, every hard drive is a disaster waiting to happen but here is my situation:

Looking at making a RAIDZ2 array with 4 3TB Western Digital Red hard drives. From what I've heard, they are made for NAS and last much longer. Plus the warranty is 3yrs whereas Seagate Barracudas dropped from 5yr warranties to 1yr warranties (which is very disconcerting).

3TB Seagate Barracuda is $150, 3TB Western Digital Red is $260

FINALLY the question...

With FreeNAS 8's Advanced Power Management Feature:

1. Do I really need a higher quality drive if I set my standby time to 10mins and power setting to "1"?
2. With a power setting at "1" will I have slower read/write speeds or is this not related?

Thanks for reading through this!!

**QUICK EDIT**
Also wanted to add that my NAS is a backup of my Drobo. Drobo has five drives, only on during a backup and is set to dual disk redundancy (I believe its the same as RAIDZ2)

 

[Stephens]

I wouldn't pay $260 for a WD Red, which is basically a Green ($100 for 2TB) with the features returned that WD previously took away from the Green drives (TLER management, ability to disable aggressive spindown, etc.), over an equivalent Seagate Barracuda @ $150. The red will also be slower as I believe the Barracuda you're looking at is 7200 rpm and the Red is basically a Green (slower spindle speed). There's no proof the WD Red will last longer than the Barracuda. Some will say it's "built for 24/7 operation" but the only person I've read who actually took apart equivalent green and red drives sees no significant hardware upgrades which would make them more reliable for 24/7 operation. You do get the better warranty with the Red. That leaves (as you allude to) power management. I'm not sure I completely understand your intended FreeNAS usage, but if you're just using it to store daily backups, you can set the drives to spin down and that'll pretty much end any significant power draw.

$150 is about where I'm expecting to get my 3TB drives. I don't really want to go 7200 rpm because Low Power drives generat less heat, virbration, and undergo less stress. But our options are limited. My prefernce is a 3TB version of the HD204UI @ $150, but it exists in the same place unicorns do. Seagate 3TB @ $150 is OK, but 7200 rpm. I don't mind using an occasional WD green, but not for the whole array. WD Red is still too new and pricey from what I've seen.

 

[cyberjock]

Looking at making a RAIDZ2 array with 4 3TB Western Digital Red hard drives. From what I've heard, they are made for NAS and last much longer.

There is ZERO evidence that they will last longer. NONE. The drives haven't been out long enough to even give a "guess" as to how well they will perform. The best evidence from people that have bought them so far is that they've worked for a month(or two.. however long they've been out). They could be complete crap or they may be the best thing since sliced bread. I really wish people would quit jumping to the conclusion that because Red is "made" for NAS because the manufacturer says so that it must be better. It may be worse for NAS. How many times have companies in all industries marketed a product being good for a particular function and the product failed miserably for that one function. Seagate used to sell consumer drives that were "designed for desktop RAID".

Until Red drives have been on the market for at last 18 months there no way I will ever take anyone's opinion of red drives seriously unless they have 100+ drives and have alot of info comparing the red to previous models. (Hint: I don't expect ANYONE to ever meet this criteria because anyone that owns 100 drives buys the Enterprise class drives!) Anyone trying to give opinions of the long term usability of Red is clearly ignorant at this point in time. Any "evidence" is strictly anecdotal from the manufacturer. Do you really think that if it sucked in a NAS that WD would come out and say "buy our great NAS drives that suck in NASes"?

Just as Stephens says Red drives are effectively Green drives with features put back in Green drives that were taken away before. Any other changes to the disk are likely proprietary WD info and we'll never know.

The only advantage the Red drives have is that "great" 3 year warranty. I put great in quotes because 5 years ago the standard for all hard drives I bought was 5 years. Now it's 1 year unless you spend more money for a "better" model or pay for a 3rd party warranty. Overall though, I'd think that companies have put more "trust" in hard drives with longer warranties. But the fact that green drives are 1/2 the price, perhaps the Red drives DO suck and WD has priced the Red drives accordingly expecting high RMAs and expect to still make a profit. Just like with all brand new products that aren't "field" tested and "field" approved.. buyer beware.

Personally, the drives I've seen last the longest are the ones that are running 24x7, well cooled with fans and never powered down regardless of model, brand, and size.

 

[jgreco]

The drives haven't been out long enough to even give a "guess" as to how well they will perform.

By that logic, you're always going to have to wait for drives to get near the end of their production lifecycle before you "trust" them. This fails to take into account that changes often happen during a product lifecycle, sometimes minor things, sometimes major, that can also impact the reliability of a particular model or set of models. And then you get screwed by EOL, when you're buying drives that are that far into their lifecycle. Bad choices.

A lot of us who work with big storage have plenty of experience working with "desktop" rather than "enterprise" grade drives. There are ups and downs to desktop drives. So let's say you have a Hitachi Ultrastar 5K3000, HUA5C3030ALA640. Those are about $320. You can get the Deskstar variant of the 5K3000 for around $160. That's literally a 2:1 price difference. The difference is that when you're paying half the cost for the Deskstars, there tend to be more failures, but they're a ton cheaper, so replacing 20% more failed drives is a cost *savings*, even if you're BUYING the replacements, and you can save the capex and upgrade to the Deskstar 5K4000's in a year or two anyways, for the same cost. We've sold storage systems consisting of a hundred or more drives to customers repeatedly with "desktop" grade drives, because in many cases, capex is an overriding concern, and in the industry we work in here, replication of data at the server level is customary (RAIS not RAID), so loss of a drive is a trivial issue.

Generally speaking, Western Digital is fairly competent, even if it pains me to say that a bit. There's no reason to guess that there's any problem with Red drives, or that the Red drives aren't simply re-flashed Greens. Some of us have been tweaking Greens for some time to de-green them anyways. WD isn't stupid, they know this. There are some basic factors here:

1) Some users would twiddle their drives but don't know how/don't want to risk voiding warranty/etc

2) Some users don't know that they could/should do that to a Green, but will pick a "NAS" drive for their NAS because that makes it obvious

3) There's some evidence that enterprise drives, in general, are more reliable than desktop drives, but also a substantial body of evidence against it. That's to be taken with a grain of salt, though, and some explanation. We've purchased plenty of super-pricey enterprise drives over the years to get 15K RPM drives, high capacity drives, etc., but we see a relatively high failure rate on them. Of the less-pricey desktop drives we buy, they tend to be 5400RPM drives. These fail much less frequently. This is probably mostly attributable to the fundamental differences in the drives - 15K drives run *hot*. Even 7200RPM drives tend to fail more often than 5400's. 5400's are pretty damn safe technology.

4) NAS has become substantially more prevalent, and with the "post-PC" era upon us, finding a way to keep their finger in the rapidly growing network-attached market is just good business sense.

5) There's money to be made. The people who would have bought a Red drive under the above points but were buying Green instead are suddenly paying a premium that they might not have been paying before.

So kudos to WD for figuring out that they can sell their Green drives as a different color, with different firmware, and make more money doing it, into a growing market.

 

[cyberjock]

By that logic, you're always going to have to wait for drives to get near the end of their production lifecycle before you "trust" them. This fails to take into account that changes often happen during a product lifecycle, sometimes minor things, sometimes major, that can also impact the reliability of a particular model or set of models. And then you get screwed by EOL, when you're buying drives that are that far into their lifecycle. Bad choices.

That is exactly my point. This new "Red" model has just been released. Nobody on the planet has any prior history.

For example, my home NAS is all WD Green drives. If I had to build another NAS today, I'd probably look around at the Green drives to see if there are any glaring problems using them in a NAS, then use them again because of my prior history with them. Reds have no history with anyone. All that we know is that they are much more expensive than a Green drive and have a longer warranty.

Is it possible that the latest generation of Green drives absolutely suck? Yes. But statistically you have a better chance of success trusting models you have used before than trying out a new model that has never really been used in the real world.

I will say that 15k drives fail so often. I went through 4 of them back in 2001-2006. Thank goodness I had a 5 year warranty so they were all free minus the cost of shipping.

 

[sonny81]

This has been extremely informative and I'm EXTREMELY grateful for all the hard work everyone put in explaining their view points.

Especially grateful to the fact that you've probably saved me a lot of money here.

Here's my current plan:

Seagate Barracudas for some reason never failed on me. I know some guys who have had TERRIBLE experiences with the Barracudas, but their the only drives I've owned that have never let me down.

So I'm going with the $150 Barracudas and just getting the 3yr extended warranty from Newegg.

Plus, all my "critical" data is obsessively saved on a Drobo, USB drive (at a friend's house in a safe), and online via my website where clients purchase prints (I'm a wedding photographer).

 

[toddos]

The only other recommendation I'd add is to either buy several drives from different places or buy them over a several week period in order to minimize the chances that all of your drives will come from the same manufacturing batch. In theory, if a batch is bad all drives from that batch are more likely to die and if your entire data set is made up of drives from the same batch you run a higher risk of losing everything rather than just one or two drives (depending on if you go with Z1 or Z2).

 

[jgreco]

That is exactly my point. This new "Red" model has just been released. Nobody on the planet has any prior history.

For example, my home NAS is all WD Green drives. If I had to build another NAS today, I'd probably look around at the Green drives to see if there are any glaring problems using them in a NAS, then use them again because of my prior history with them. Reds have no history with anyone. All that we know is that they are much more expensive than a Green drive and have a longer warranty.

Actually, it's looking likely that the Red drives are the Greens with some firmware tweaks, although WD has made some claims about a new mechanical balancing technology for vibration reduction. The warranty and price differences are not technical issues and don't affect reliability, so may not be relevant. From a technical angle, then, it's unlikely that WD - a major manufacturer with an established history - made significant changes to produce this drive. Since the hardware platform is exceedingly similar to the Green drives, probably with fewer differences between Red and Green than between generations of Greens, there's no particular reason to be overly suspicious. You're getting all obsessive over what is primarily a marketing ploy that has a few firmware tweaks and maybe a minor hardware variation to create it.

Do you want to just run out and get the latest model of any hard drive? Mmm, possibly not. Risk. Yes... risk. But that applies at all levels. New model? Risk. New firmware revision? Risk. New sourcing for components to make existing model? Risk. Someone sneezed at the factory? Risk. Forklift operator dropped the pallet? Risk. You don't really get to understand, or even know about, all the risks. To be certain, it is good to favor vendors with a good history. It's good to favor product lines with a good history. But when you start talking about requiring 18 months to validate a drive, it's likely that you're no longer able to get the drive that you just finished validating... the manufacturer has moved on to a slightly newer version, with some more incremental changes, etc., and you need to start your 18 month validation all over again, because you once again have a drive that "nobody on the planet has any prior history" with.

It sounds to me like you're fixated on the color change aspect of this, but from an engineering perspective, it really sounds like this is mostly a marketing thing, where WD has given us some basic NAS-favoring features that people have either been manually hacking on in the Green series drives, or just doing without. I don't get the paranoia you're exhibiting.

 

[jgreco]

The only other recommendation I'd add is to either buy several drives from different places or buy them over a several week period in order to minimize the chances that all of your drives will come from the same manufacturing batch. In theory, if a batch is bad all drives from that batch are more likely to die and if your entire data set is made up of drives from the same batch you run a higher risk of losing everything rather than just one or two drives (depending on if you go with Z1 or Z2).

Buying drives from different manufacturers (getting harder, that) is also a fine idea, though you want to be careful to acquire units with similar characteristics if at all possible. This idea horrifies some drive buyers, especially those who've been indoctrinated by FE's for NAS/SAN vendors (who would prefer to avoid having to deal with the increased potential for firmware bugs in a diverse population) and manufacturer reps who of course would not want you buying a competitor's product for any reason... but in this day of 1:1 SATA bus topologies and ZFS's software-centric design, it's actually quite a reasonable thing to do.

sonny81's comments about Barracudas are particularly striking to me, since we're an OLD Seagate shop. Those old Barracuda ST12550N's were the poster child for dying in the most spectacularly inconvenient ways. We picked up some of the latest gen Barracuda 3TB's and had 50% DOA/infant mort on those as well.

 

[cyberjock]

Actually, it's looking likely that the Red drives are the Greens with some firmware tweaks, although WD has made some claims about a new mechanical balancing technology for vibration reduction. The warranty and price differences are not technical issues and don't affect reliability, so may not be relevant.

That "new mechanical balancing technology.." is stuff that I'm referring to. How "tested" is it? It would suck to buy 16 of these, put them in a NAS, then find out that they're failing left and right because that new mechanical balancing technology is causing premature failures.

Yeah, there's lots of risk just between different firmwares, I got bit with the Seagate CC1H firmware issue from 2008 time frame. But when they start applying new technology that should be a "wait and see" approach for anyone investing big dollar signs.

Check out Intel's track record with their products. Remember the SATA port snafu from Sandy Bridge? Yeah, mistakes get made. Intel has one of the most rigorous R&D testing facilities in the world when it comes to testing electronics. Somehow that one made it through. I'd hate to be on the receiving end of 16 drives with some nasty bug that is a hardware issue. What do you think WD is going to do if they figure out later that they have a hardware issue? Will they try to brush it under the rug(like Seagate did with CC1H)? Will they try to preemtively fix the problem(like Intel did with the Sandy Bridge SATA ports)? You'd be giving too much credit to WD if you think they're going to preemptively RMA a batch of bad hard disks. How much money will YOU lose to shipping RMAs for 16 drives? Will you lose your zpool because you had 3 drives fail while trying to RMA a zpool of bad disks?

There's a reason why you shouldn't buy the first year of a new model car. Statistically they have alot more issues than others. I prefer to take the very cautious stance of letting other people spend their money on them and see what happens. I'd rather not drop $2k on hard drives to find out I made a horrible mistake. Maybe I'm being overly conservative. But when it comes to my data(and my family's car) we'll stick with waiting a few months for the Red to be field tested(and stick to 2nd year+ vehicles).

I wouldn't say I'm fixated on the color change aspect but more on the "intended audience" and the engineering behind that "intended audience". This is the first time any company has pushed a "NAS" drive like this, and I'll be cautiously optimistic about

Do I expect the Red drives will perform fine and have an acceptable lifespan? Yes. Am I a little paranoid about just trusting a new model because a company says it's great for a given use? Perhaps. Am I about to drop $2k on an untested model? Heck no.

I'm sure neither of us are going to convince the other of their opinion, and I'm completely okay with agreeing to disagree. It's all about thought process. I've been bit with screwups before by companies(last one was Seagate in 2009) and I'll never be so naive to think that if a company says it will work then it surely will again. Try buying $2500 worth of hard drives and then 4 months later having to replace them ALL because drives are failing out because Seagate had an undocumented "feature" that made RAID controllers think that the drives had failed. I saw 2 drives fail one night with no load at all during the night. They sat in a box for 2 years because I wouldn't trust them and nobody else I knew would trust them either. No RMA from Seagate because they passed diagnostics. Yes, the drives and the controller were both on the "manufacturer's compatible list". Same firmware and everything. So pardon me if I'm a little skeptical of a new model.

 

[Joshua Parker Ruehlig]

The Reds look interesting to me but at $260/3TB I just can't justify them over a $100/2TB HD204UI. I just had a HD204UI fail on me and it really didn't mean too much stress cause I'm running RAIDZ2. Just gotta switch it out then, scrub and should be all good.

 

[jgreco]

That "new mechanical balancing technology.." is stuff that I'm referring to. How "tested" is it? It would suck to buy 16 of these, put them in a NAS, then find out that they're failing left and right because that new mechanical balancing technology is causing premature failures.

Sure, but you can say that about ANY feature, hardware change, firmware change, etc. The whole point I'm trying to make is that waiting 18 months to make sure there aren't "problems" with a model is untenable; I know people who've built arrays that started experiencing large numbers of failures around 2 years, or 3 years, etc. anyways, so there's nothing magic about 18 months. It's just taking you out far enough that you're pretty certain not to be able to get the hardware model and revision that you just finished "validating", which seems counterproductive.

I'm also not saying that paranoia isn't good. You should be aware that the possibility of failure exists. However, mitigation needs to take place at many levels. Don't get drives from a single manufacturing batch. Don't get drives from a single manufacturer, for that matter. Build with more redundancy rather than less (two RAIDZ2 vdevs consisting of 8 data/2 parity, rather than one RAIDZ2 vdev consisting of 18 data/2 parity, yes, you lose some space). Don't entrust all your data to a single NAS. etc. It's great to be sensitive to changes in the hardware, but at some point you have to get your feet wet.

 

[jgreco]

The Reds look interesting to me but at $260/3TB I just can't justify them over a $100/2TB HD204UI. I just had a HD204UI fail on me and it really didn't mean too much stress cause I'm running RAIDZ2. Just gotta switch it out then, scrub and should be all good.

Yeah, the pricing looks like it's intended to gouge the NAS market a bit. I don't mind paying a price premium if there's a good reason, but the benefits are kind of marginal in this case.

 

[cyberjock]

Sure, but you can say that about ANY feature, hardware change, firmware change, etc. The whole point I'm trying to make is that waiting 18 months to make sure there aren't "problems" with a model is untenable; I know people who've built arrays that started experiencing large numbers of failures around 2 years, or 3 years, etc. anyways, so there's nothing magic about 18 months. It's just taking you out far enough that you're pretty certain not to be able to get the hardware model and revision that you just finished "validating", which seems counterproductive.

I'm also not saying that paranoia isn't good. You should be aware that the possibility of failure exists. However, mitigation needs to take place at many levels. Don't get drives from a single manufacturing batch. Don't get drives from a single manufacturer, for that matter. Build with more redundancy rather than less (two RAIDZ2 vdevs consisting of 8 data/2 parity, rather than one RAIDZ2 vdev consisting of 18 data/2 parity, yes, you lose some space). Don't entrust all your data to a single NAS. etc. It's great to be sensitive to changes in the hardware, but at some point you have to get your feet wet.

There's a big difference between a firmware update and a new technology being applied. Sure, both can turn your hard drive into a paper weight, but adding a new technology is a MUCH riskier adventure than a firmware update.

 

[jgreco]

If you summed up the number of "new technologies" that have been "added" to hard drives over the years, you'd be asking yourself why these things aren't miniature Space Shuttles and able to fly you to the moon and back. Realistically, "new technologies" are often trivial and trite changes, often from other existing product lines.

As for hardware versus firmware, quite frankly I think you're just wrong. I can definitely think of many more "firmware" based flaws in drives in the past decade than I can fundamental hardware flaws. It's fine to be wary of hardware flaws, but ... geez.

 

[sonny81]

So with 7200rmp desktop drives will the "Advanced Power Management" settings affect read/write speeds or is this just a "better power utilization" feature? I have my drives set to 1 and am using CIFS (I know I know) but just want to make sure I'm not getting more loss in performance than gain in power efficiency.

 

[jgreco]

Mostly kills read/write speeds when you're doing a read/write after a long period of inactivity where the drive has gone into a lower power mode, and done something like parked the heads or spun the drive down - like you could see a 5-10 second pause while the drive gets back up to operating speed, then it's all good after that.

You can see HGST talking about it here in combination with an Adaptec RAID controller, though the general effects on HDD access speed are going to be valid without the Adaptec controller...

http://www.hgst.com/tech/techlib.nsf/techdocs/5DF05E0FB03F8B4D86257515000384C4/$file/Hitachi_Adaptec_PowerMgmt_WP_final.pdf