6TB hard drives
- Thread starter Bozo
- Start date
Chewy509
Wotty wot wot.
Anandtech review of 6TB drives: http://anandtech.com/show/8263/6-tb-nas-drives-wd-red-seagate-ec-hgst-he
I only have a couple of 6TB drives the standard Seagates, which work just fine though are not in an array. Are they crippled some way for RIAD 5?
CougTek
Hairy Aussie
More because of the time it takes to rebuild a RAID5/6 array combined to the possibility of an unrecoverable parity error during the array rebuild following a drive failure. The possibility of losing the array after a drive replacement, particularly for a RAID5 array, is non-negligeable.
In the case of the WD Red, if the 6TB model is like its smaller siblings of the same family, it is rated to do one unrecoverable error per petabyte (10e15). A 5 drives RAID5 array of WD Red 6TB drives has 30TB of data to write when building the array. Therefore, during the array formation, it has a one on thirty-three chance of doing an unrecoverable error (1 PB/30TB), resulting in the lost of the array's integrity. There is also the possibility that a second drive failure occurs during the many hours needed to repair a damaged array.
RAID6 is less problematic since it can sustain two drive failures or unrecoverable errors before losing the array, but it still isn't ideal.
Using drives that are rated to not do more than one unrecovable error per 100PB (10e17), like most SAS drives, diminishes the unrecoverable error problem by a factor of 100.
In the case of the WD Red, if the 6TB model is like its smaller siblings of the same family, it is rated to do one unrecoverable error per petabyte (10e15). A 5 drives RAID5 array of WD Red 6TB drives has 30TB of data to write when building the array. Therefore, during the array formation, it has a one on thirty-three chance of doing an unrecoverable error (1 PB/30TB), resulting in the lost of the array's integrity. There is also the possibility that a second drive failure occurs during the many hours needed to repair a damaged array.
RAID6 is less problematic since it can sustain two drive failures or unrecoverable errors before losing the array, but it still isn't ideal.
Using drives that are rated to not do more than one unrecovable error per 100PB (10e17), like most SAS drives, diminishes the unrecoverable error problem by a factor of 100.
Bozo
Storage? I am Storage!
Do the Red Pro drives do any better?
In terms of unrecoverable error rate, the WD Red pro drives in 2TB, 3TB, and 4TB are listed as being 1 per 10e15 according to WD in their FAQ doc. I don't know if there is a WD Red pro 6TB.
Well, somebody is buying the high-grade Seagate and Hitachi 6TB drives for $600+, so I would assume they are used in some sort of arrays that are reliable enough. The WD Red drives are the SOHO NAS type.
CougTek
Hairy Aussie
RAID1 and RAID10 are perfectly fine. Some people are also clueless and buy these drives to put them in cheap RAID5-configured NAS.
There are some SAN that have additional parity check, like the Fujitsu Eternus DX and probably the HP 3PAR 7x00 I have at the office, which can use huge capacity drives in RAID5 without issue. Those are not cheap though (typically >20K$).
And to answer Handruin's question : there currently is no WD Red Pro above 4TB.
There are some SAN that have additional parity check, like the Fujitsu Eternus DX and probably the HP 3PAR 7x00 I have at the office, which can use huge capacity drives in RAID5 without issue. Those are not cheap though (typically >20K$).
And to answer Handruin's question : there currently is no WD Red Pro above 4TB.
jtr1962
Storage? I am Storage!
One of the comments to the article mirrors my thoughts here:
"I know some will disagree,.. whatever - but what is the goddamned POINT of 7200rpm platter disks for consumers nowadays? Seriously. If you care about performance you either have an OS/Apps SSD or a hybrid drive even."
"I know some will disagree,.. whatever - but what is the goddamned POINT of 7200rpm platter disks for consumers nowadays? Seriously. If you care about performance you either have an OS/Apps SSD or a hybrid drive even."
Bozo
Storage? I am Storage!
Do they make a 4TB or 6TB SSD?
time
Storage? I am Storage!
Maximum capacity is only 4-6 times different, but cost is still an order of magnitude.
As an analogy, if a gasoline car costs $20,000, are you prepared to pay $200,000 for an electric car?
As an analogy, if a gasoline car costs $20,000, are you prepared to pay $200,000 for an electric car?
Is the point to suggest that 5400 RPM is fine? I'd still like reasonable access times when fetching my data especially in a much larger drive. Until someone comes up with consumer level fully automated storage tiering solution why not try and get the fastest spindle you can when 6TB of SSD isn't likely affordable.
jtr1962
Storage? I am Storage!
The point is huge drives are mainly for bulk storage of things like video where access time doesn't matter, and STR of even a 5400 RPM drive is more than sufficient. The downsides of 7200 RPM over 5400 RPM are heat, noise, lifetime, reliability, and density. Before SSDs became affordable faster spindle speeds made sense, at least for the boot drive. Now they really don't but drive manufacturers haven't caught on to that fact yet.
I've probably purchased my last mechanical drive when I bought a 2 TB Samsung three years ago for three reasons. One, it seems drives reliability is going down drastically as densities continue to increase. Two, 5400 RPM drives don't seem to be readily available. Three, while multiterabyte SSDs aren't yet affordable, if 3D flash scales as promised they will be by about the time I will start to be concerned about my 2TB drive aging out.
jtr1962
Storage? I am Storage!
Again, the point is that the stuff people typically put on 4TB or 6TB drives doesn't require the fast access times of an SSD. Fast access times are only important when you have lots of small files, not a few large ones. That being the case, why not make decrease spindle speeds and increase capacity/reliability? Heck, why not 3600 RPM or even slower? I'll bet most users of large drives wouldn't notice the difference.
timwhit
Hairy Aussie
There are still many computers sold with one large mechanical disk. With all the crapware that comes on these system, it makes sense to lower access times.
jtr1962
Storage? I am Storage!
Yeah, but we're talking here about huge drives which aren't going to be on computers sold with one mechanical disk. Just looking at a sampling of systems on Newegg typically those one disk machines have something between 500 GB to 1TB. I don't care if 7200 RPM drives are made for those who may still see some advantage to them, but I do care that if I'm in the market for a hard disk for whatever reason I have practically no choices for anything slower. Most of the slower choices are WD, and most of those are the Intellipower crap. It still seems HDD manufacturers think sticking a higher RPM number on the package is going to help them sell disks. That's no longer true. For many like me, the opposite is true. I don't want 7200 RPM. I don't want the noise, the reduced reliability, the higher power consumption/heat. 5400 RPM drives or slower run virtually silent and much cooler. Given that SSDs have reached the point where even a cheapskate like me will use nothing else for system drives, I want any mechanical drives in the system to not add any noticeable noise, nor do I want to have to deal with making sure they have sufficient airflow. 7200 RPM drives are mostly a relic of another time which is thankfully gone as far as I'm concerned. Hopefully in a few years the same will be true of spinning disks in general.
You don't want Intellipower (5400 RPM), but you also don't want 7200 RPM? :scratch: I guess that leaves you with a 5900 RPM Seagate 6TB drive if they ever release one.
I suspect that the 6TB drives are designed for servers first and the cheaper versions trickle down to the lower market.
I don't think most of the 6TB market is geared toward people downloading **** from the interwebs and playing it on their HTPC.
The Seagate 6TB 7200GB drive is only $300, which is not that much more expensive than their 4TB 7200 RPM drive. You can buy a Seagate 4TB 5900 RPM drive that is a decent performer, though the 6TB 7200 RPM is noticeably better.
Mercutio
Fatwah on Western Digital
I want large consumer-class disks that are still capable of operating properly in an array. Buying 7200rpm drives seems to be the most straightforward way at the moment to ensure that. If the slower drives weren't deliberately crippled, I wouldn't have a problem with drives being slower, but don't forget that plenty of people still need fast + capacity and very few of those people have a tiered storage infrastructure (ZFS/ReFS et al) right now.
Mercutio
Fatwah on Western Digital
Honestly, I'm not sure what the general use case is for huge drives. Video storage and centralized backups are about the only things I can think of. Big single drives are a huge PITA to back up and even virtual machines do better if they're given a spindle to themselves rather than having to compete for I/O with all the other guests on the same host. I'm sure the motto is that if they build it, someone will buy it, but nothing short of video production or downloading mind-blowing amounts of **** from the interwebs is going to touch that level of disk usage.
I had a similar problem with the arrays at EMC when they'd rotate them out for our group every 6-12 months. I'd always have to make the argument that total capacity was not as important as IO management on spindles (i.e. I want more spindles of smaller size rather than less with larger size). When ordering the new arrays they always focused on the great capacity and up-talked getting 4TB NL_SAS (AKA SATA) when I was happier with 600/900GB 10K SAS drives. If I make a RAID 10 LUN with 8 x 4TB drives, 90% of it goes unused because performance is shit with too many VMs on a LUN. Things got better once they introduce caching LUNs with SSDs but still wasn't ideal.
I'd be fine with companies building 3.5" SSDs and packing in 4-6TB of NAND even less efficiently. Are the chips really that expensive or is the market just holding out and cashing in on the faster tech? I mean, even if they gave us a 4-6TB SSD with a less than optimal controller is still eons better than a 6TB 7200RPM HDD even for general storage.
jtr1962
Storage? I am Storage!
Yep, that would be exactly what I think the market would need instead of 7200 RPM huge mechanical disks. Get the worst grade flash, even the stuff that might get trashed because it's way below specs, the shittiest controller going, don't even try to pack it in efficiently, and you'll still end up up with 4 to 6 TB SSDs in a 3.5" form factor which would run circles around anything with spinning disks. While it would cost more, I'll bet such a drive could come within a factor of two of the costs of a mechanical drive. Would I buy something like a 3 to 4TB 3.5" SSD for maybe $300 right now, even if the access times were 1 ms (which is glacial for anything solid state)? Sure I would, if they existed. No noise, little heat, slow by SSD standards but still an order of magnitude faster than spinning disks. Well worth paying twice as much per GB for if you ask me.
And no, I don't think the chips are that expensive. This seems to be a case of the market just milking people. Keep mechanical disks around, and people are willing to pay a huge premium per GB for SSDs. Start making huge SSDs with what would probably otherwise be throw-away NAND, and now you can't charge as much even for fast, 2.5" SSDs. As I said earlier, 3D NAND looks to change the game anyway regardless of what the mechanical drive manufacturers want within about 3 or 4 years. I think things could happen even sooner if we made huge, but crappy (by SSD standards anyway) SSDs. Even if a 6TB SSD was only good for a few tens of write cycles, that may well be enough given how they might be used by most people. If you're downloading and saving music or videos, chances are good you're just writing a lot of the cells only once or twice, and then just reading them from that point forwards.
It will be years yet before a 6TB flash drive is $300 and by then a 6TB HDD will be closer to $100. Nobody will make a cheap SSD with really shitty flash memory. Well maybe they can resurrect OCZ.
