Seagate releases 8TB 3.5inch drive

[zoggy]

http://www.kitguru.net/components/hard-drives/anton-shilov/seagate-ships-8tb-customer-development-hard-drives/

 

Seagate Technology on Thursday said that it had shipped the first samples of hard disk drives (HDDs) with 8TB capacity to its customers. The company did not reveal any peculiarities regarding the product or the name of its clients.

 

 

[sacretagent]

wow 176 TB in a unraid server

parity check for 2tb in my server is about 11 hours x 4 to get to 8 TB = 44 hours  aka 2 days parity check ...

 

 

guess downloading the internet is possible in the future

[bfeist]

http://www.theregister.co.uk/2014/08/26/seagates_eight_terabyte_spinner/

 

[c3]

http://www.seagate.com/about/newsroom/press-releases/Seagate-ships-worlds-first-8TB-hard-drives-pr-master/

[Fireball3]

amazing 

[Owel]

how long will preclear last? at least 3 cycles of course!

[Fireball3]

at least double of 4TB = 44+ hours 

[SSD]

Will be intetesting to see WD's response.

 

Seagate has this shingled disk technology (SMR), but I do not think WD has yet developed anything to compete. Seagate may have the size advantage for quite a while. The biggest disk draws a lot of sales by large data centers, and WDs 6TB was not in the biggest disk category for very long at all.  I expect that 8T drive will but downward price pressure on the 6T quicker than WD expected.

[garycase]

It will indeed be interesting to see if WD responds to this in the near future (clearly they will eventually).

 

Note that Seagate has not released the technical details of this drive ...i.e. # of platters; whether it's a PMR or SMR based drive; etc.    One reason WD has (so far) avoided using SMR platters is that this technology has lower performance than the PMR platters, since entire blocks must be re-written for every write.

 

In any event, the capacity "wars" are clearly continuing ... although I doubt there will be much price pressure on WD's 6TB drives from this particular drive.    It's currently selling for $66.38/TB ($531) ... with a 30-day warranty !!  But just it's existence will undoubtedly be a thorn in WD's side that may cause some interesting movement in the hard drive market.

 

 

[c3]

The 8TB is an SMR drive, but it is untrue that every write requires blocks to be re-written. SMR drives avoid the re-write situation just like SSDs avoid it, as much as possible, with the on device controller hiding the process.

 

I suspect SMR drives will be a good fit for many unRAID users, those with growing, static content.

[garycase]

Shingled writes absolutely require rewriting the entire "shingled section" when you do a write.  They minimize the impact of this by not "shingling" the entire platter -- it's done in relatively small groups of "shingles", so only the current section requires rewriting through the end of that section ... and by using a fairly large buffer, so they can often buffer the "next" write while the previous section is being completed.

 

But it's absolutely true that SMR drives cannot match PMR drives in performance due to this requirement.

 

Note that despite Seagate's pronouncements last year of the density improvements that SMR would provide, their 6TB units do NOT use this technique.    While I agree it seems likely the 8TB units do, I have not seen any confirmation of this.

 

 

... The 8TB is an SMR drive ...

 

Every press release and review I've seen notes specifically that Seagate has "... declined to clarify the platter count, spindle speed, and whether the 8TB unit uses shingled magnetic recording technology."   

 

Where did you see confirmation of the recording technique?

 

I agree, however, that despite the write speed penalty of SMR, it's a good fit for users with primarily static content, where the bulk of the activity is reading.  For those applications, the higher density possible with SMR technology easily offsets the slower writes.

 

 

 

[WeeboTech]

Interesting read for those wanting to learn about SMR.

http://www.seagate.com/tech-insights/breaking-areal-density-barriers-with-seagate-smr-master-ti/

[garycase]

Interesting read for those wanting to learn about SMR.

http://www.seagate.com/tech-insights/breaking-areal-density-barriers-with-seagate-smr-master-ti/

 

Yes, that does a nice job of showing why SMR writes are slower than PMR ... and also shows how Seagate is "banding" the groups of shingles so there's a reasonable limit to how much needs to be re-written for every write.

 

[c3]

First, the re-write only occurs on a re-write, not first write (or complete band write).

 

This article gives a very clear picture of when re-write is required. If the content is never changed (aka static), no re-write will occur.

http://www.seagate.com/tech-insights/breaking-areal-density-barriers-with-seagate-smr-master-ti/

 

Second, in the worse case, the entire band is not re-written, only the required portion. Because if you write the entire band, again, no re-write is required.

 

The firmware is all about creating the entire band write situations, avoiding the read/write/re-write situations.

 

Seagate has been shipping SMR since the 5TB ST5000DM000 and Kinetic drives http://www.seagate.com/about/newsroom/press-releases/shingled-magnetic-recording-milestone-pr-master/ Like a year ago...

 

Now, everyone can start setting noatime, etc...

 

 

[garycase]

First, the re-write only occurs on a re-write, not first write (or complete band write).

 

Actually, any write to the drive requires this.  A "first write" by a user is a rewrite of the existing information -- the drive doesn't "know" that there's nothing already on the tracks.    How much needs to be re-written depends entirely on the size of the "bands" that are shingled.  If, for example, the bands are 64MB, then to write anything on that band, the drive needs to read the entire 64MB, update the data for the area of the write, and then re-write the entire band.    Many articles on SMR say this makes SMR drives undesirable for applications that require a lot of random writes.    The Seagate article clearly shows this "banding" as the solution to minimize the number of tracks that need to be rewritten ["... all data in surrounding tracks are essentially picked up and as a result will need to be rewritten at a later time ..."]

 

 

If the content is never changed (aka static), no re-write will occur.

 

Of course -- if the content is static then nothing's being written to the drive, so clearly there's nothing to re-write.  Reads don't cause any changes 

 

 

... Seagate has been shipping SMR since the 5TB ST5000DM000 and Kinetic drives ...

 

They have certainly been issuing press releases about the technology and advertising how many drives they're shipping with it ... but they are very silent about just WHICH drives use the technology.    According to the technical reviews of their 6TB enterprise units, these do NOT use SMR technology ... apparently because of performance issues.    As I noted above, I suspect the 8TB units DO use SMR, but I've seen nothing to confirm that, and Seagate is completely silent on the issue, saying only that "... more details will be announced later."    It's strange that they're totally silent about this technology if in fact that's what the drive is using. 

 

They clearly list SMR as a feature with their 5TB ST5000DM000 drive in its specs ["... SMR Technology adds greater than 25% capacity growth ..."], so I'd certainly expect them to note in on drives where they use it.  It is NOT listed as a characteristic of their enterprise class 5TB units or any of their 6TB units ... and as I noted earlier, the technical reviews of these drives also note that they do NOT use SMR.

 

 

In any event, the question I asked was pretty simple:  What is your basis for the following comment?  (Or are you just assuming it uses SMR?  ... I assume that as well, but I do NOT know that, or any of the other detailed characteristics of the drive.)

 

... The 8TB is an SMR drive ...

 

[c3]

"When a user needs to rewrite or update existing information, SMR drives will need to correct not only the requested data, but any data on the following tracks". You can find it just below Figure 2. on http://www.seagate.com/tech-insights/breaking-areal-density-barriers-with-seagate-smr-master-ti/

 

I know the 8TB drive is SMR because I asked when I got them.

[BRiT]

Who did you ask?

[zoggy]

http://www.storagereview.com/seagate_8tb_hdd_now_shipping

[garycase]

http://www.storagereview.com/seagate_8tb_hdd_now_shipping

 

Like all other press releases and tech details I can find on these drives, this simply confirms that Seagate isn't saying what the underlying technology of the drives is.  e.g. "... As to other details, Seagate isn't providing any."

 

The Storage Review article DOES say "... An educated guess however would suggest this could be the next iteration of Seagate's SMR ..." -- and I agree with that;  but an educated guess does not equal actual confirmation of the technology.

 

[zoggy]

some additional articles about it... the one on extremetech gives a decent overview of the different techs involved in creating more dense drives.

 

http://www.computerworld.com/article/2599020/cloud-storage/seagate-scoffs-at-helium-ships-world-s-first-8tb-hard-drive.html

http://www.extremetech.com/computing/186624-seagate-starts-shipping-8tb-hard-drives-with-10tb-and-hamr-on-the-horizon

[garycase]

Yes, I'd already seen those.

 

As with virtually all other articles/releases about the new drives, the article notes that although Seagate has made a lot of noise about the potential benefits of SMR, "... actual data on which drives actually use SMR is hard to come by."

 

Seagate is VERY silent on this issue.  The only drive they're actually selling where the specs on the drive actually SAY it's an SMR drive is the 4TB ST4000DM000 drive.  Note that the subsequent 6TB drives do NOT mention SMR, and are notably faster drives.    Although a few folks have speculated that the 6TB drives were also SMR drives, Seagate has NOT said that;  does NOT list SMR in the specs; and MOST tech sites think the 6TB units are NOT SMR drives -- to quote Extreme Tech: "... For the 6TB drive, we believe Seagate used conventional perpendicular magnetic recording (PMR) rather than shingled magnetic recording (SMR) or heat-assisted magnetic recording (HAMR)."    Apparently the issue is performance -- as Extreme Tech noted, "... SMR is theoretically a lot slower than PMR when it comes to writing data ..."  => but there's not much empiral evidence to confirm that or to show how well Seagate might have been able to mitigate that by using relatively small bands of shingles.

 

One would certainly think that if these new drives are SMR drives that Seagate would be loudly proclaiming this in their press releases instead of being completely silent on the technology they're using.

 

 

[c3]

I think you mean ST5000DM000.

[garycase]

I think you mean ST5000DM000.

 

No, I meant the ST4000DM000.    But it's true that the ST5000DM000 has the same comment in its specifications:  "... •SMR Technology adds greater than 25% capacity growth by maximizing the number of tracks per inch on a single disk."

 

... so I should have listed both of those drives as units that Seagate includes SMR in their specifications.

 

It's very interesting that they do NOT mention this technology in any of their subsequent drives' specifications.

 

[Helmonder]

Sorry to resurrect an old topic..

 

But would this work as a parity drive ?  Or would it slow down the whole system... I am waiting for the WD 8TB RED but that seems to take for ever..

[pkn]

Sorry to resurrect an old topic..

 

But would this work as a parity drive ?  Or would it slow down the whole system... I am waiting for the WD 8TB RED but that seems to take for ever..

SMR drive (8 TB Seagate HDD) - normal read, slow write.

PMR drive (most other drives) - normal read, normal write.

 

So yes, using an SMR drive for parity would slow down any and all writings to unRAID (read speed should not be affected).

 

I personally plan to use two PMR-type 4TB drives in RAID-0 for parity. This would give me 8 TB "fast" parity drive, and allow using of 8 TB data drives in array. And, at the same time, would allow to avoid slower writing to other, PMR-type data drives, in array (slowing of writing to SMR data drives seems to be unavoidable).

 

Of course, it does not matter if array consists of only SMR-type drives.