Slashdot Mirror
6 Terabyte Hard Drive Round-Up: WD Red, WD Green and Seagate Enterprise 6TB
MojoKid writes The hard drive market has become a lot less sexy in the past few years thanks to SSDs. What we used to consider "fast" for a hard drive is relatively slow compared to even the cheapest of today's solid state drives. But there are two areas where hard drives still rule the roost, and that's overall capacity and cost per gigabyte. Since most of us still need a hard drive for bulk storage, the question naturally becomes, "how big of a drive do you need?" For a while, 4TB drives were the top end of what was available in the market but recently Seagate, HGST, and Western Digital announced breakthroughs in areal density and other technologies, that enabled the advent of the 6 Terabyte hard drive. This round-up looks at three offerings in the market currently, with a WD Red 6TB drive, WD Green and a Seagate 6TB Enterprise class model. Though the WD drives only sport a 5400RPM spindle speed, due to their increased areal density of 1TB platters, they're still able to put up respectable performance. Though the Seagate Enterprise Capacity 6TB (also known as the Constellation ES series) drive offers the best performance at 7200 RPM, it comes at nearly a $200 price premium. Still, at anywhere from .04 to .07 per GiB, you can't beat the bulk storage value of these new high capacity 6TB HDDs.
I've seen in general, three lines of HDDs. Basic desktop/laptop drives, premium desktop/laptop, and enterprise grade drives which are designed to all wind up at the same firmware level to minimize issues when in RAID controllers.
However, a "NAS" hard drive? Is this something a step down from enterprise drives, but designed for a device like a Drobo, or some other solution that really doesn't care about background drives, uses RAID 5 or 6, and expects drives to blow out over time?
Are the Red drives designed to be paired or run in RAID arrays specifically, as opposed to the Green line that is made for power savings?
Awfully long summary to say "you can haz 6TB HD"
Is anyone with significant amounts of data not caching their frequently accessed data on SSD? Rotational is still about 8x cheaper than SSD these days, but the days of rotational speed for cold data are numbered. Storage is easily abstracted so it's not a legacy concern. A lot of shops I know have already invested in a complete switchover to full-SSD (we're talking racks of SSD) with tape backup.
Even my home file server uses two tiny second gen 64gb SSDs for read/write caching for ~20TB of data. I just buy the cheapest, biggest rotational drive whenever I start running out of room. When the price on those new Seagate 8TB drives (currently $230) drops to under $150 I will probably start swapping out my oldest 2TB drives to avoid having to upgrade the case in this decade.
moox. for a new generation.
Aren't both of those the bottom of the barrel these days?
You'd be nuts to trust your porn stash to a 6TB consumer drive right now. Buy two 4TB drives, and back that stuff up. Give the 6TBs a year or so to see if there are any reliability issues with these capacities, and for the price to drop a bit.
I don't build a machine these days that doesn't have mirrored hard drives. You realistically can't backup 6TB worth of data, so barring some horrible FS failure (which is rare these days in Linux land) your best bet is RAID1.
Yes Francis, the world has gone crazy.
Fastest: Seagate.
Best Warranty: Seagate.
Best Cache: WD Red....or the Seagate...the article conflicts between the first two pages.
Cheapest: WD Green.
Seagate notables: Full drive encryption available at a firmware level. AF and Legacy disks are separate models.
WD Red notables: 5400RPM spindle speed.
WD Green notables: None - nothing distinguishable from the Red drive, except a shorter warranty.
Sandra Benchmark results:
Seagate: 167W/168R.
WD Red: 138W/138R.
WD Green: 133W/133R.
Atto results are shown on a messy graph with no clear numbers, but Seagate wins that benchmark as well (albeit with a closer delta).
HD Tune Pro results basically reflect the transfer rates from above. Seek times for the Seagate are 11ms for both write and read, with the WD Red having a 16/17 set of scores and the WD Green being less than an integer higher. Burst rates are again better on the Seagate (276R/304W), with the WD Green being 217/220 and the Red being 217/218.
Crystal mark, basically the same numbers.
Futuremark, prettier graphs with wonderful titles like "video editing" and "importing pictures", with the results a closer race, each drive having its own task at which it wins (even the green). Not much different from the 3TB numbers, and not that much different from each other.
There were no mentions of reliability metrics; presumably none of the disks failed during benchmarking. Consult your usual biases and experience regarding which drive is likely to fail or not - this was strictly a benchmark review, and shockingly, the enterprise-grade drive with the highest rotational speed and biggest cache that costs the most money got the best score.
If you can't figure out what he meant from the context then I think you might want to re-evaluate who the worthless fuck is.
Sorry, teleporters just kill you and then make a copy. A perfect, soul-less copy.
7 words back of that is "$200" ...so the astute mind would assume $ as the unit referenced.
Pretending this is my office full of bitter coworkers..
Wear-leveling with SSDs isn't about reliability, more about longevity because a specific bit (really page) on a chip can only be rewritten so many times. Though once that point is hit, the data is still readable, just not writeable. Not really the same scenario as the head of a HDD crashing, making data-recovery far more laborious and expensive.
Reading is hard.
Tic-Tac-Toe, Global Thermonuclear War, and relationships all have the same winning move.
Is SSD reliability/longevity still lower than the HDD's, or is that no longer true (or is the opposite true nowadays)?
http://hardware.slashdot.org/s... -- Two petabytes of data written on a cheap, consumer-grade drive and it's still going strong. Not all drives last that much, but basically you can use the drive for two decades without much worry. Of course you're still better off using a non-SSD for something that keeps on chugging data to the drive 24/7, but for pretty much any other use-case they're perfectly fine these days.
Having experienced SSD failures.. NO you cant read from them. SSD drives do a catastrophic failure, you do not get a chance to read from them before full failure, they just do a complete fail and all data is gone forever.
Do not look at laser with remaining good eye.
Samsung's 850 Pros have a 10-year warranty, although they are still quite expensive.
Also techreport.com has been running an endurance test, and a couple of the drives have reached 1.5 petabytes of writes without failing. I think they all lasted well beyond the manufacturers' expected write limits.
Basically, they've reached the point now that the average consumer can't wear out a drive.
1. I suspect even the author couldn't tell you whether it's .04 to .07 cents or dollars per GiB.
2. By my math it's $279/6144=$0.05 to $479/6144=$0.08 per GiB, not $0.04 to $0.07.
3. Why are we using GiB when hard drive capacities are expressed in GB/TB?
Any sufficiently unpopular but cohesive argument is indistinguishable from trolling.
Hey, the S looked like it was crossed out, OK?
Bert
Spindle speed does not matter much, the number of spindles and temperature does. So WD Red is the best choice.
I wonder how those Seagate 8TB Shingled Magnetic Recording (SMR) perform in a NAS. 33% more capacity for the same price.
Well his own username does indicate that he's scum....
I think that depends on the nature of the failure. The flash media itself will fail as described, but that fact will be almost completely hidden by a decent drive controller with a storage reserve - by it's nature the failure is easy to recover from, at least until such time as a sizable percentage of the drive capacity has so failed.
On the other hand it sounds like the SSD controllers tend to be less reliable than on a HDD, and if the controller goes you get an immediate catastrophic failure. It doesn't matter if the data is still theoretically safe on the flash media, if it can't be accessed short of re-soldering the circuit board.
--- Most topics have many sides worth arguing, allow me to take one opposite you.
+1, there is no dd if=/dev/brokenDisk of=prayer.img with a failed ssd.
My experience has also been a zero warning, complete loss with SSDs.
And you'd trust your data to a first-gen drive technology? Backups are great and all, but it's still a hassle. I've been burned by enough cutting edge hardware already, I'll let the IT departments deal with the teething pains. I'll be waiting for at least v1.1, maybe 2.1, before I'm tempted by a marginal up-front cost reduction.
--- Most topics have many sides worth arguing, allow me to take one opposite you.
Marketing weasels.
I've seen the same for hard-drives, and tend not to have content that would be worth paying for recovery, so for people like me there's not much of a difference in reliability. Only one out of 5 of my SSDs has failed so far, which is a much lower rate than I've seen with hard drives in my array.
I'd love to know how they can be "high reliability" drives when they're new tech and haven't been tested out in the real world for an extensive period of time. "High reliability" is only something you can demonstrate retroactively, or with a proven technology.
Irony: Agile development has too much intertia to be abandoned now.
Once the electrons are out of the gate, the data is -gone-. No amount of recovery is going to do the job, ever.
This is my biggest concern with SSDs. Yes, they can have a longer MTBF, but when they go, they take your data with it, making backups more imperative.
The ironic thing? Since SSDs make the need for backups that much more urgent [1] We have far fewer tools for backup than we did on PCs 20 years ago (when an average user could get a desktop tape drive, a ZIP drive, removable SCSI hard disk, or other media.) For non-enterprise backups, we have external hard disks, USB flash drives, and offsite file servers [2]. Even optical drives are becoming uncommon. External hard disks and USB flash drives are not archival media. They -might- hold their data, but are not warrantied for it.
It would be nice if some company could make an appliance that did a disk-to-disk-to-removable-media appliance. The backup program would copy data to the device, and data would stay on a set of RAID protected HDDs, as well as eventually copied to removable media [3]. A bare metal restore would be easy -- if the appliance is connected via USB, have it present a DVD-ROM with the OS or recovery software. If on a LAN, have a USB flash drive or image that would get a machine booted enough to find the appliance and start a restore.
[1]: With HDDs, a recovery from a format isn't too difficult. SSDs usually follow up a format with a TRIM command, zeroing (or more exactly, writing 1s) to all the blocks, either right then, or as the drive feels like it. "Unformatting" a SSD is pretty much impossible with a modern OS that does proper TRIM commands. Add a decently smart encryption system like BitLocker that zeroes out the sectors with master volume keys multiple times, and it can almost be assured that a delete or a format results in data forever gone.
[2]: Cloud storage seems like a working idea, but it can take a good while to fetch lost documents and rebuild the entire OS and machine. With a local backup solution, most backup programs offer a simple bare-metal restore, no Internet access needed. There is also the fact that a machine needs to have the OS, updates, and the cloud provider's software loaded and logged in before a restore can happen. Having the OS local means a complete bare metal restore is a "press 'restore' and walk off" action.
[3]: Tape comes to mind. The main advantage of tape (or offline media in general) is that some hacker who gets access to the SAN controller can't just purge all media with a single command. A lot of companies have excellent replication of SAN data, but that replication will happily replicate the "delete everything, including all snapshots" as well. Plus, tapes can be physically set read-only where only a reflash of the tape drive could allow the cartridge to be written to. I wish someone could make a consumer level tape drive, perhaps using a SSD as a buffer to prevent shoe-shining. There is a Thunderbolt based tape drive for Macs by mTape for $3699. If someone made a product like this (but a price more palatable to consumers) that could tolerate USB 3 (or maybe even USB 2), and work well under Windows, Linux, and other operating systems, they might have a best seller. Especially with the fact that intruders now have moved from just accessing data to actively modifying and destroying it, so backups are even more crucial than they were before this year.
In fact, I'd say that with the ease data is permanently destroyed, a consumer level backup appliance might be quite a seller.
Would much rather have RAID6 of 5 2TB drives. Basically would rather have the most drives in the biggest RAID that would allow the lowest price per gigabyte...
I beg to differ.
Anecdote: I have a stack of 4 Seagate ST325082A 250GB PATA drives I bought in 2006 with the specific intention of building a RAID. That RAID is still running, almost uninterrupted since November of that year (breaking for fan replacements and the odd power cut), with no failures.
Political debates have me rolling my eyes so much I think I got optical whiplash. I should sue. - Foamy The Squirrel
I purchased the first 7200rpm disk available to consumers nearly 20 years ago now. The WD Expert, 18gb if I recall.
http://www.prnewswire.com/news...
I've always hated the performance of disks, big enthusiast primarily because I knew it
was the biggest bottleneck by far.
Fast forward to today and I am utterly bamboozled why people continue to purchase the bastard things. I detest them. They run hotter, cost more, are slightly more likely to fail, are noisier and the performance difference is utterly negligible.
You need only look at any older SSD review, where they include 3x5400rpm disks, 4x7200rpm disks and 5 SSD's - the graph is difficult to read because even the fastest hard disk is vastly slower than the slowest SSD.
We're entering the age now where even mid-to-basic level nerds have a NAS in the home. I'd wager a reasonable portion of people have SSD's in their main machines / laptops and some 'big dumb storage' in the rest of the house, be it USB storage or a NAS. :/
YET HOWEVER,.... when I went to buy new disks last year, in a new size range (5TB) do you think I could find the 5400's? Nope, the fucking 7200's were the first available. Infact this trend has gone on for a few years now. You used to get 5400's in the new size first, then when the tech slightly improved, they'd do the 7200rpm model. This no longer seems the case.
I actively DON'T want 7200rpm disks in my bloody NAS (which is now locked inside my kitchen cupboard, with a fan on it and ventilation door open to keep the damn noise low) My disks managed to hit 57c (134f) because I couldn't find god damn 5400rpm disks, hence the new fan install
7200's are pointless, it's like buying premium grease for the axle of your horse and cart. :/
If you want performance, SSD, if you want space? Big, dumb, slow, cool, quiet 5400rpm disk. If you want to piss away money, 7200rpm disk. Bastard things.
Really wish I didn't completely need to buy disks when I did, only 12 weeks after I got mine, the WD red and greens were out to buy
In conclusion, avoid Toshiba 7200rpm disks, they are not only hot, the bloody spindle motors are noisy to boot.
I've always relied on density improvements of HDDs to save me when my disks fill up rather than getting rid of junk or archiving museum pieces. Nice to finally see densities improving again as there is only a couple hundred GBs remaining. Only thing that will suck is having to move to GPT/EFI.
Have a rule only a single pair of mirrored disks shall go into a machine and only software raid shall be used based on years of getting burned by worthless hardware controllers and parity based raid schemes.
Machines are only rebooted to install security patches and suspended to ram when not in use. Disk I/O is not a factor at all... more than enough RAM available to cache everything needed to get work done in a daily basis with no noticeable delays of any kind.
Personally so far SSD costs too much, does not offer competitive densities, not suitable for swap partitions, lack deterministic erasure and lingering reliability fears stemming from lower maturity level of memory controllers and supporting electronics.
Drives that old are pretty irrelevant in a discussion about multi-terrabyte drives. While I do have a single one of those notorious 1.5TB Seagates, all of it's siblings died a long time ago.
Seagate has done quite a bit lately to earn it's bad reputation.
A Pirate and a Puritan look the same on a balance sheet.
BINGO
The underlying issues with flash can be and are successfully hidden by the controllers in modern SSDs for most workloads (very heavy write loads can be problematic) but that hiding comes at a price. The firmware in a SSD is far more complex than an a HDD and so for a given level of engineering effort it will be less reliable. In particular i've noticed corruption after unclean shutdown to a far greater extent on SSDs than HDDs.
note: i'm known as plugwash most places but i screwd up registering that here somehow in the past and now can't register
IMHO serious archives go on tape. However you have to be very serious about it since a couple of hard drives is a lot cheaper than an LTO6 drive and a few tapes - tape doesn't win until you hit large volumes and long timescales.
A ten year old tape you pull out of a box is going to work apart from a tiny fraction of a percentage of the time. A drive - not so likely since the spindle lubricant doesn't last forever and polished surfaces stick via diffusion. A twenty year old tape should have been transcribed years ago but is going to work unless it has got hot or damp in storage. A thirty year old tape is probably brittle and needs to be read with care, but I've sent a couple of dozen off to be transcribed. It was seismic data so file formats that could handle a few bits missing here or there, and errors outside the file headers have little impact due to "stacking" multiple datasets that overlap. However those reels from the early 1980s and late 1970s preserved effectively all the data put on them despite less than ideal storage (a shed in a humid subtropical climate).
Hard drives are not designed to last for a decade in a box. A decade powered up is ironicly likely to result in less dead drives than powered off on a shelf. Tapes don't have to deal with high speeds and are instead designed to last. They die from the substrate getting brittle over decades, the oxide peeling off the tape over decades and magnetised zones on one section of tape magnetising an area on the next loop of tape, once again over decades.
All that said, if you only have 6TB or so to keep, and you don't want to go for a pile of Blueray disks, getting a couple of drives every few years (3? 5? 7?) is a lot more sane than mucking about with tapes.
They have had a few dud models that are not like those drives from 2006 or even 2000.
D2D2Cloud is the most common method for home backup, that's how Mozy and Crashplan work, they backup to both a local HDD and to the cloud storage, if you have a non-catastrophic loss you restore from local media, if you've lost it all you restore from the cloud, either over the wire or if you have a lot of data you pay for optical or HDD recovery. Crashplan adds the option to backup to a friend or family member, if you design it so that they are using a separate drive for your target you can remove that drive and recover from localish media without paying. I'm personally using Crashplan and backing up to three targets (local, brother, and crashplan central) as well as acting as a target for my father-in-laws free Crashplan backups (he backs up locally and to me, no need to pay for a subscription since I have a machine that's on 24x7 and he has a fairly small volume of data to be backed up).
There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
The underlying flash won't fail as described. Data stored on flash degrades over time. It's an analogue medium (just like a disk and anything else in reality above quantum levels) that is then quantised. When a cell is brand new, the difference between a 1 and a 0 is big and easy to detect. Over time, without rewrites, the difference lowers. When you rewrite a cell, you get a big difference again, but usually a slightly smaller difference than the previous time that it was rewritten. The controller knows roughly how the data degrades over time and will avoid using cells that don't look as if they can hold their content for a year (don't store SSDs on the shelf unpowered for a few years and expect to be able to read anything back!) and will refresh cells periodically. Once a cell fails to meet that requirement immediately after a write, it will be discarded and reused for something else (there was a nice paper at EuroSys this year about using the less-reliable cells for things).
If a given cell won't hold its value in a readable way after an erase, there's a good chance that it won't before the erase either (and the erase may have been triggered by a refresh because it's getting close to the borderline). If a single cell that is visible to the user is starting to exhibit these properties, it means that the wear levelling has run out of reliable cells to use and the device is basically dead.
I am TheRaven on Soylent News
I have a 40MB Seagate drive that was still working happily 12 years after it was bought (no idea if it still works - I haven't tried to use it for ages). Like your anecdote, mine tells you nothing at all about modern Seagate drives.
I am TheRaven on Soylent News
to remind you: GP did not specify how old the drives were he was referring to. Only that they were Seagate. I'm not the only one on this thread who proved his claim to be a pile of shit.
Political debates have me rolling my eyes so much I think I got optical whiplash. I should sue. - Foamy The Squirrel
The ironic thing? Since SSDs make the need for backups that much more urgent [1] We have far fewer tools for backup than we did on PCs 20 years ago (when an average user could get a desktop tape drive, a ZIP drive, removable SCSI hard disk, or other media.
Desktop tape drives were shit and their tapes often unreadable after just short periods. Unless you bought a used exabyte drive, your tape drive at home was certainly shit during that era. Zip drives never held whole HDDs worth of data, and you may have forgotten that they cost $15/100MB, they are probably the most expensive removable media ever (not counting stacks of platters from vintage DASDs, which are not conveniently swappable) after the original MO drives. We can still use removable hard drives, there are many solutions for that. And we still have "other media". Your complaint is nonsense. HDDs are so cheap now you can just buy a second HDD, back up to it, and then disconnect it — in comparison to the rest of the PC, it's cheaper now than it's ever been and also cheaper now than backups have ever been before.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
HDDs are of course faster at the beginning and slow towards the end. Not one graph in several pages of benchmarks showed how these drives compared after the first few gigabytes of storage.
Waterfox - a Firefox fork with legacy extension support, security updates and better privacy by default.
And how many of these drives protect against bit rot? Protection be built in as standard with Reed-Solomon error correction (magic afaik) or similar.
Waterfox - a Firefox fork with legacy extension support, security updates and better privacy by default.
Reason GiB is used instead of GB was that GB would denote 1024MB, as opposed to 1000MB (and likewise for MB and KB). In the semiconductor memory space, that was fine, but for HDs, manufacturers wanted to mean MB as 1000, as opposed to 1024, and that confused the market. That's why KiB, MiB, GiB et al were all invented. It's a good idea to use them, since lay people ain't gonna start calculating capacities in powers of 2.
.
I have always wanted a data 'black hole' that I could retrieve data from. But it still isn't there. One that does automatic HSM (hierarchical storage management) so you store in on fast devices, it stays there a while, then migrates (automagically) to slower devices, and eventually to 'archival storage' that can be slow to get to.
So far I haven't found an answer I can afford (personally). -- If you know of something, please let me know! --- Think 'net to SSD, to Disk, to slow disk/nas, to tape or optical drives. Tape and optical data still needs to be read and written on occasion to stay fresh (especially tape). Tape also wares out (so do optical media after 50 or so years, tape degrades dramatically after 5). -- also need multiple copies for when one gets 'bit rot' happens.
Commercially I like IBMs Tivoli Storage Management (just because I used it), but that comes at a pretty hefty price, but it works well when set up and tuned correctly.
... "When you pry the source from my cold dead hands."
Wear-leveling with SSDs isn't about reliability, more about longevity because a specific bit (really page) on a chip can only be rewritten so many times. Though once that point is hit, the data is still readable, just not writeable.
This sometimes happens in the best behaved drives, but in practice that's more of a myth than something that really happens. Most SSDs start forgetting data after they're powered down when they hit their end of wear life. See the SSD endurance at TechReport for some real world examples.
Maybe, but their addressing & data ain't tied to powers of 2 like semiconductor memory is. Which is why you have odd number of GB, like 500GB or 250GB or so on. Not 512GB or 256GB. Hence this trend to GB and GiB. I'm glad that this separate terminology exists - it was annoying when KB meant both 1024 AND 1000.