Seagate Shipping 8TB Hard Drives for Testing

[Guest]

Seagate reports key client feedback on testing units has been very positive.

Seagate Shipping 8TB Hard Drives for Testing : Read more

 

[iceclock]

nice

 

[CaedenV]

It is truly amazing to see HDD technology attempt to retain relevance right now. I mean, 8TB is nothing to scoff at; It is an absolutely mind-blowing amount of storage for spinning-platter media. I mean I am still impressed by the 3TB drives that I picked up last year.
But the fact of the matter is that this is a dead technology. In the enterprise sector the pricing for high quality HDDs is not that much different from SSDs anymore. SSDs also provide ridiculous amounts of speed, more graceful/predictable failures, much less heat output, much less power usage, and typically a much smaller form factor for a relatively small increase in price. Plus SSDs are starting to ship with insane 10 year warranties... I don't think we will be seeing HDDs with those kinds of claims any time soon.

As cool as this tech is, it is bound to be one of the last new products that they make. Even Seagate is jumping into the SSD market these days, so it should not be too long now before we see them start replacing their own enterprise equipment with SSD based solutions.

all the same, I really wonder what kind of tech they are using for these. Even at a 1.2TB/platter density it would need some 7 platters to hit 8TB, and I am not sure you can cram that many layers in a standard HDD form factor.

 

[iceclock]

id like to see 3 tb ssds

 

[thundervore]

Toms, fix your comment system. nothing is more aggravating that writing a lengthy comment then when you try to post you get the message "ERROR_CANNOT_FIND_DOC_ID" then the page refreshes causing you to lose the whole comment!!!!!

 

[dgingeri]

4TB drives have been out for two and a half years, and we're just now testing to 8TB? It's gonna be at least another year before they get to stores. This is very slow progress. Back in the pre-GB days, drives were doubling in size every year. Even in the pre-1TB days, we were doubling every 18 months. This is pitiful progress. Are they even trying these days?

 

[InvalidError]

But the fact of the matter is that this is a dead technology. In the enterprise sector the pricing for high quality HDDs is not that much different from SSDs anymore.

There are tons of archival situations were cost per byte is far more important than speed or latency. In those situations, spinning disks still have a 5-10X cost advantage over SSDs. Decent quality HDDs do not fail that often when treated correctly unless you get a defective unit and today's lowest-power 7200RPM 3.5" HDDs operate in the 5-8W range, which is not that bad compared to 3-5W active power from SSDs.

The only HDDs SSDs might kill any time soon are those overpriced 10-15k RPM monstrosities.

As far as reliability goes, if your data is really important, you should be using RAID6 or some other forms of logical and physical redundancy regardless of whether you choose to use HDDs or SSDs anyway.

 

[InvalidError]

Back in the pre-GB days, drives were doubling in size every year. Even in the pre-1TB days, we were doubling every 18 months. This is pitiful progress. Are they even trying these days?

Approaching physical limits has that effect on technology... the same thing happened to lithography: we used to move down one process node every ~18 months, then it turned to 24 and now, with 14nm (or 32/28nm to 20nm for TSMC, GF, UMC and friends), we are around 30 months.

Product cycles and progress are not going to get any faster any time soon if ever; all signs point towards things slowing down even further unless there are some major breakthroughs in technology, processes and materials.

 

[CaedenV]

4TB drives have been out for two and a half years, and we're just now testing to 8TB? It's gonna be at least another year before they get to stores. This is very slow progress. Back in the pre-GB days, drives were doubling in size every year. Even in the pre-1TB days, we were doubling every 18 months. This is pitiful progress. Are they even trying these days?

What, no /s at the end of your comment? HDD tech has already broken several 'impossible' barriers, and each one becomes more and more difficult to pass. Meanwhile flash based drives are eating away at the cash-cow enterprise market, and are getting more and more affordable all of the time while offering higher density and lower power solutions. HDD tech is absolutely mind-boggeling in how it has lasted this long. You cannot expect it to continue that same rate of progress forever.

Plus there is the whole access issue. Who is going to buy an 8TB HDD and only want ~150-200MB/s access to it? On a sequential read you are talking about almost 12 hours, and with non-sequential loads you are talking about ridiculous amounts of time to access data. It becomes more and more like drinking the ocean through a straw. It is still OK for things like 4K video storage, but once you start adding multiple users who demand this kind of storage then you very quickly find that it is just not enough on the throughput end of things, no matter how much storage there may be on the drive.

 

[InvalidError]

Plus there is the whole access issue. Who is going to buy an 8TB HDD and only want ~150-200MB/s access to it?

Someone who needs to archive terabytes if not petabytes of data that does not need to be accessed that often... like Netflix's video archive, Facebooks' photo/video archive, Google Play's app/games/content archives, etc. Any such large archive will usually have multiple RAID arrays capable of reading and writing at over 500MB/s each so each server with 40Gbps LAN connection would be able to handle about eight of those arrays.

If you need to serve a subset of that data on a more regular basis to a wider audience, that's what tiered storage systems and caches are for... pull something from the archive at 500MB/s, replicate it across content caches and serve it out at several GB/s aggregate bandwidth. A single SSD would not be able to handle this alone either.

 

[CaedenV]

But the fact of the matter is that this is a dead technology. In the enterprise sector the pricing for high quality HDDs is not that much different from SSDs anymore.

There are tons of archival situations were cost per byte is far more important than speed or latency. In those situations, spinning disks still have a 5-10X cost advantage over SSDs. Decent quality HDDs do not fail that often when treated correctly unless you get a defective unit and today's lowest-power 7200RPM 3.5" HDDs operate in the 5-8W range, which is not that bad compared to 3-5W active power from SSDs.

The only HDDs SSDs might kill any time soon are those overpriced 10-15k RPM monstrosities.

As far as reliability goes, if your data is really important, you should be using RAID6 or some other forms of logical and physical redundancy regardless of whether you choose to use HDDs or SSDs anyway.

Is there really a market for consumer 8TB HDDs? I just assumed (perhaps incorrectly) that drives this size would only have a real use in more enterprise style solutions where the price per GB is closer to $0.20-.30/GB rather than consumer drives where data is a mere $0.05/GB. SSDs may still be more expensive than that, but in another year or so when these drive hit the market SSDs will have made up most of that difference while offering a host of other benefits (including reliability which HDDs just can't offer at those sizes).

The other big supposed use for these kinds of drive is surveillance so that you can store several months, or even a year of video on a drive. But this is Seagate, King of the 2 year warranty. Are we really going to see 2 year drives being thrown into appliance settings that may have lifespans of 5-10 years? How out of balance is that?

About the only use for these things is cold storage where you have a massive collection of movies that you don't watch all that often, but don't want to get rid of. And even there I would like to think I have a decent movie collection (granted I am not a big movie buff) and can easily fit all of my movies in ~1.3TB of space at native quality. I suppose I would need more space if more of my collection was in bluray, but even if it was we would be talking about ~2-3TB of storage... not 8TB.

I guess I am just confused as to who actually buys these things in enough bulk where it could be cheap enough to sell as a consumer drive, which is why I assumed it would be more expensive.

 

[dgingeri]

But the fact of the matter is that this is a dead technology. In the enterprise sector the pricing for high quality HDDs is not that much different from SSDs anymore.

There are tons of archival situations were cost per byte is far more important than speed or latency. In those situations, spinning disks still have a 5-10X cost advantage over SSDs. Decent quality HDDs do not fail that often when treated correctly unless you get a defective unit and today's lowest-power 7200RPM 3.5" HDDs operate in the 5-8W range, which is not that bad compared to 3-5W active power from SSDs.

The only HDDs SSDs might kill any time soon are those overpriced 10-15k RPM monstrosities.

As far as reliability goes, if your data is really important, you should be using RAID6 or some other forms of logical and physical redundancy regardless of whether you choose to use HDDs or SSDs anyway.

It's not just in the enterprise where HDD's cost per byte comes up. I have 980GB of stuff I have in storage (drivers, isos, utility programs, music, movies, and other stuff) that really doesn't need a SSD for storage. I'm sure as heck not going to put that out on a cloud storage. It would take days to copy. I don't access it that much, but I want it around when I need it. So, I have my system set up with a SSD for the OS and office programs, one for games so I don't lose my game settings when I rebuild my system every couple months, and a 3TB HDD for file storage.

Also, I happen to be a systems admin for an enterprice backup device/ software company. We have dozens of systems that use slower HDD based storage in raid arrays for long term storage. Our main profit is from older tape libraries, but we have been moving to disk based backup systems for the last several years. I just built a system with 288TB of storage with deduplication tech (which would basically storage 2880TB of data in about 2/3 of a rack) for testing. Many companies are still using this type technology rather than backing up to the cloud. Most companies can't afford internet pipes to have a cloud backup be their primary backup, so cloud backups aren't going to be the primary backup storage, let alone primary storage, for quite a while yet. A great many companies are still on old tape libraries for their primary backups instead of moving to disk backup systems with tape being secondary.

HDDs will be around for a very long time. Heck, even tape is still going to be around for at least another decade. I figure HDDs will be around for another 50 years before it is supplanted by SSDs. I'm just really disappointed that HDD manufacturers have consolidated so much and slowed down the research and improvement rate so much.

 

[sea monkey]

But the fact of the matter is that this is a dead technology.

137M shipped in Q1 of 2014

http://www.statista.com/statistics/275336/global-shipment-figures-for-hard-disk-drives-from-4th-quarter-2010/

In the enterprise sector the pricing for high quality HDDs is not that much different from SSDs anymore.

"At a $/GB measure, SSDs have dropped from upwards of $3/GB back in 2005 and 2006 down to as low as $0.67/GB for specific models in 2012, but they are still far above the cost of HDDs, which can be lower than $0.09/GB for some Western Digital and Seagate models. Even if it is assumed that SSDs continue to fall at their historical rates and HDDs fall at their historical rates, SSDs will still be about $0.15/GB in 2020 ($0.06 more than HDDs are today), while the HDDs vendors are promising may be as low as $0.03/GB."

http://www.enterprisestorageforum.com/storage-hardware/ssd-vs.-hdd-pricing-seven-myths-that-need-correcting.html

As cool as this tech is, it is bound to be one of the last new products that they make.

"WD Demos Future HDD Storage Tech: 60TB Hard Drives"

http://www.tomsitpro.com/articles/wd-hamr-hdd-heat-assisted-magnetic-recording,1-1396.html

 

[InvalidError]

Is there really a market for consumer 8TB HDDs? I just assumed (perhaps incorrectly) that drives this size would only have a real use in more enterprise style solutions where the price per GB is closer to $0.20-.30/GB rather than consumer drives where data is a mere $0.05/GB.

When is the last time you looked at 7200RPM SAS drive prices? I'm not finding much in your $0.20-0.30 range.

Looking at no-frills 7200RPM SAS 3/6/12G drives, there are plenty of 3-6TB units around $0.10/GB. That would be more in-line with what a sane person would want to use for a near-line enterprise-class archive.
http://www.newegg.com/Product/ProductList.aspx?Submit=Property&N=100007603%20600003485%20600003482%20600510654%20600003340%20600484470%20600003344%20600486291%20600003489%20600457700%20600361769&IsNodeId=1&name=2.5TB%20-%203TB

Looking at 10k RPM and higher, prices jump to a more drastic $0.50-$1.00/GB.

 

[SirGCal]

Plus there is the whole access issue. Who is going to buy an 8TB HDD and only want ~150-200MB/s access to it? On a sequential read you are talking about almost 12 hours, and with non-sequential loads you are talking about ridiculous amounts of time to access data. It becomes more and more like drinking the ocean through a straw. It is still OK for things like 4K video storage, but once you start adding multiple users who demand this kind of storage then you very quickly find that it is just not enough on the throughput end of things, no matter how much storage there may be on the drive.

Let's learn a bit more about how everything works...

112MB/sec is all the throughput you're going to get with a fully saturated 1Gb/sec network connection. (it's the bits vs bytes issue, HDD/SDDs and windows transfer rates are calculated in Bytes (8 bits), networks are measured in bits).

Plus, those of us using these in other situations get signifigantly faster speeds. I run 8 drives in a RAID 6 configuration (or RAIDZ2) and they give roughly 600MB/sec througput either way which puts the pain to my 10G/sec internal network which most people, especially home users, don't even have/use. However a home user using their 1G network (or even more likely, slower still wifi network) is going to hit that limit before the HDD speed limit in many cases, even on a single drive. So multiple users hitting it won't matter unless they are slamming the pants off of it constantly in which case the network will be another limiting factor.

While I'm all for faster and faster drives, we need faster network interfaces first. My home is wired CAT6E with using a consumer grade wired router can sustain 1G/sec no problem. Most people don't have their homes wired and use Wifi which is much slower overall (though it's catching up fast, finaly). Still using consumer grade equipment, the speed of the drive isn't your issue.

 

[mapesdhs]

Toms, fix your comment system. nothing is more aggravating that writing a lengthy comment then when you try to post you get the message "ERROR_CANNOT_FIND_DOC_ID" then the page refreshes causing you to lose the whole comment!!!!!

Tip: before clicking Submit, highlight all your test and press CTRL+C (assuming
you're using Windows). That way, if the post barfs, you can get back to the
posting page again and CTRL+V, or paste it into an editor, etc.


Re HDDs, one thing that does bug me about newer models is the percentage
capacity increase has become rather low over existing models. It's quite a bump
from 2TB to 4TB, but from 6TB to 8TB is less interesting, and as some have said,
backing up that data is a pain. Would be nice if one could get hw like LTO6 at a
decent price, but such tech will always be expensive because the target market
can afford the pricing. Roll on LTO7/8, but they'll still cost a fortune.

Using multiple HDDs in some kind of mirrored RAID is one solution, but that's
not going to help in the event of a fire or nasty electrical fault that messes with
everything. When I was a sysadmin at a research centre, the DLT backup
tapes were stored long-term at the other end of the building, in a big fire-proof
safe, so even if we had no power at all, we could always take recent tapes
elsewhere to extract the data. A company using just RAID1 or whatever in the
same unit as the main server is asking for trouble.

I agree with CaedenV, the I/O rate of newer rust spinners has to be significantly
improved for them to be practical IMO. Manufacturers love to quote sequential
rates, but as soon as one copies real data, the bw is nowhere near that high.
I have 15K SAS drives that have long exceeded the sequential rates being
quoted here, but they suffer from the same issues; somewhat less so with the
faster access times, but still nowhere near an SSD.

Ian.

 

[dgingeri]

Re HDDs, one thing that does bug me about newer models is the percentage
capacity increase has become rather low over existing models. It's quite a bump
from 2TB to 4TB, but from 6TB to 8TB is less interesting, and as some have said,
backing up that data is a pain. Would be nice if one could get hw like LTO6 at a
decent price, but such tech will always be expensive because the target market
can afford the pricing. Roll on LTO7/8, but they'll still cost a fortune.

Fun little facts from someone in this industry:
1. Building a LTO6 drive, only including the physical build, not counting engineering costs involved, costs roughly twice what it does to build a 4TB SAS hard drive. The engineering costs behind the LTO6 were over $12million. (In other words, we going by very slim margins for current LTO6 pricing.)
2. It would actually be cheaper to build an external RAID controller, a 12 drive external SAS frame with SFF-8088 connectors and port multiplier, and 12 2TB drives than it would for a small tape library with 1 LTO6 drive to handle 12 tapes. It's even cheaper to just build 12 3TB USB 3.0 drives.
3. The point where it becomes cheaper to use LTO6 tapes and a tape library rather than hard drives is at 150 tapes per drive. In other words, you have to be backing up >225TB of information weekly. (My company has done much research on this matter lately.)

So, LTO6 is just not going to work at the desktop. One would have to have a significantly large enterprise level storage system to get any savings out of a tape library system, or already have much of the infrastructure in place and just upgrade to the current levels.

For desktop backups and data storage, hard drives are massively cheaper, and it is not just because the market for tape will bear the cost. It is because the economics of tape technology lends itself to be useless at the small scale and only pays back at very large scale.

My company also looked into producing a 8 bay USB 3.0 to removable SATA setup that would act like a tape library for small businesses, where small businesses could just swap out laptop SATA drives in special cases instead of tape cartridges for their weekly off site backups. They determined that it was actually cheaper and less trouble to just use USB 3.0 hard drives and specialized software to move backups off site.

 

[Benevolence]

Having worked as a master control operator for a few TV stations, we used to have these huge rooms filled with DAT tape "robots", and it worked pretty well for years and years. Once the government-mandated HD/TV switch happened (so they could resell our broadcast bandwidth to the cell companies grrr...) almost every station ditched their DAT archives and upgraded to RAID racks. Turns out that's cheaper in the short-run with faster specs, but those racks generate an awful lot of heat which required additional A/C units to be installed and operated. This is just for archive, stuff that's not going to air in the next 6 days. The power costs are significantly higher but the RAID racks gave unexpected advantages like online HD editing capabilities, FTP hosting, faster low-res previews, near instantaneous archival and other little things that made us wonder why we ever put up with that slow robot solution.

 

[Benevolence]

Also if these drives are being distributed for testing... PICK ME!!! PICK ME!!!!

 

[mapesdhs]

dgingeri, I guess really what I was getting it was that it's just a pity there isn't a high-capacity tape
solution aimed at ordinary consumers, who wouldn't need a lot of the enterprise features of a typical
LTO6 unit. Consumers are stuck with disk mirroring solutions, which really aren't that good much of
the time for various reasons.

Ian.

 

[BulkZerker]

Their biggest customer? NSA

 

[InvalidError]

Consumers are stuck with disk mirroring solutions, which really aren't that good much of the time for various reasons.

At $2200-5500 for the LTO6 drive alone, even if they made a consumer version for $1000, this is still enough to buy half a dozen 3TB drives and the LTO6 tapes to match that capacity would be cost enough to afford four more 3TB HDDs... at $160 per 3TB HDD and $60 per LTO6 tape and a hypothetical $1000 consumer-grade LTO6 drive, you break even around 36TB.

 

[mapesdhs]

I didn't mean a _consumer_ version of LTO6, I meant an entirely different tape tech aimed specifically at consumers.
LTO6 is, from the ground up, not suitable. Btw, I don't for a moment believe the production costs are really that high.
Markups on tech tend to be huge, with pricing set to what the market will tolerate. It has little to do with what
something costs to make. LTO6 is expensive, like every other pro tape tech, partly because it can be.


BullZerker, also consider NASA. Various ongoing projects acrue enormous amounts of data every day, while just
one vehicle launch generates massive quantities of video (very high res cameras).

Ian.

 

[Benevolence]

I'm just waiting for glass-plate holograms to come back into fashion for data-storage, then just maybe I'll get my very own memory crystal... I've often wondered why opto-chemistry hasn't yielded anything better than XDCam. Given that the parity is built in to the hologram (all parts reflect the whole, like a fractal!), it would seem like a very good option for data storage. I know GE has a "holographic data disk", but that's just multi-layering since it doesn't use beam interference to inscribe the data.
Just wait fellas, maybe someday 7x4ft sheet of holographic film could record your physical body on a near nano-meter scale, with enough data parity for human replication.
http://en.wikipedia.org/wiki/Holographic_memory

 

[InvalidError]

I didn't mean a _consumer_ version of LTO6, I meant an entirely different tape tech aimed specifically at consumers.

Same difference: how many people are going to bother with a slow and fragile tape-based solution that costs $50 for 2.5TB plus drive cost if they are unlikely to ever own more than 3-4 tapes? LTO6 achieves its 2.5TB per tape by recording across over 2000 tracks on a nearly 1km long tape. You are going to need some serious engineering if you want to retain LTO6's data density and most of its reliability in a consumer tape drive.

The adoption rate would likely remain very limited and if you divide the R&D cost by low shipping volume, you get high unit cost even if tape drive and tape manufacturers decided to dig in their margins to offer cheaper consumer products.

If you spend millions in R&D then ship only a few thousand units, that's over $1000/unit of engineering cost overhead to cover on top of the manufacturing and material costs.

Niche products cost a fortune partly because that's the price the market is willing to bear but also because that's the price manufacturers want to cover the risks they take by trying to cater to such niches in the first place.