Seagate’s first shingled hard drives now shipping: 8TB for just $260

[jakea333]

Since we are discussing NewEgg shipping...

 

If you have an AmEx, you can get a FREE ShopRunner membership.  That gives you free TWO day shipping.  Also, Ive probably place 20 NewEgg orders since I signed up, and they have always shipped the same day as my order.

 

Unfortunately, Newegg dropped ShopRunner from their shipping options recently (they seem to be pushing "Premier"). It was one of the reasons I used them as my go to over Amazon and other competitors. Now, they don't have that advantage over Amazon for me any longer. TD still accepts SR, but they've always been more picky about which products it applies to.

[c3]

I don't really think that the parity reliability matters much. Or that individual data drive reliability matters much.

 

You should probably look into those numbers again. Even with your alien craft, are you not going to restore? In doing so you'll perform the massive amount of data transfer to likely encounter the projected 1x10^14 URE. Yeah, restores fail too, especially when you have no regard to the reliability of the drives you are using. I doubt Gary is going to suggest you reduce the reliability of your drives.

 

[pkn]

I don't really think that the parity reliability matters much. Or that individual data drive reliability matters much.

 

You should probably look into those numbers again. Even with your alien craft, are you not going to restore? In doing so you'll perform the massive amount of data transfer to likely encounter the projected 1x10^14 URE. Yeah, restores fail too, especially when you have no regard to the reliability of the drives you are using. I doubt Gary is going to suggest you reduce the reliability of your drives.

 

I'm not suggesting to reduce the reliability of our drives... even if I knew how to do that. 

 

What I'm suggesting is that enterprise-grade drives which cost three times the price of desktop greeny, are not worth it in unRAID environment.

 

Failure of the second drive during rebuild falls under category "two-drives failure", and, as I keep repeating after more experienced unraiders - backup is the only real protection against that.

[c3]

I'm not suggesting to reduce the reliability of our drives... even if I knew how to do that. 

 

Using RAID 0 is cutting your drive reliability in half (or more).

[pkn]

I'm not suggesting to reduce the reliability of our drives... even if I knew how to do that. 

 

Using RAID 0 is cutting your drive reliability in half (or more).

Oh my... I didn't realize that, but your are undeniably right.

 

I'll make a note to myself: it's OK to use RAID-0 for unRAID cache (or anything your are prepared to loose anytime), but don't use it for unRAID parity or any unRAID array drive unless you understand the risk and ready to tolerate it.

[SSD]

I'm not suggesting to reduce the reliability of our drives... even if I knew how to do that. 

 

Using RAID 0 is cutting your drive reliability in half (or more).

Oh my... I didn't realize that, but your are undeniably right.

 

I'll make a note to myself: it's OK to use RAID-0 for unRAID cache (or anything your are prepared to loose anytime), but don't use it for unRAID parity or any unRAID array drive unless you understand the risk and ready to tolerate it.

 

If either physical drive fails the logical RAID0 drive is still recoverable. The more serious risk is if the RAID0 parity failed in the middle of a rebuild of a failed disk. I am not tracking this as a serious risk but it is something to consider. I buy very reliable HGST drives and drive failure in my array is very very rare.

[garycase]

I'll make a note to myself: it's OK to use RAID-0 for unRAID cache (or anything your are prepared to loose anytime), but don't use it for unRAID parity or any unRAID array drive unless you understand the risk and ready to tolerate it.

 

Actually, I'd be far more inclined to use RAID-0 for the parity drive than for the cache.    If one of your RAID-0 drives fails, and it's the cache drive, then everything you've written to UnRAID that hasn't yet been moved to the protected array is lost.    If it's the parity drive, then your parity "drive" just failed ... and you can recover by simply replacing the failed physical drive and doing a rebuild of the parity drive.    If you're concerned about the increased failure likelihood of a 2-drive "parity drive", then use 4 drives in a RAID 0+1 configuration.    But I don't think that's necessary.

 

... Note, by the way, that using a RAID-0 as parity is NOT "... cutting your drive reliability in half (or more)."    It DOES double the probability of a drive failure, but that's not the same as cutting the reliability in half.    If you have a drive that's 99% reliable, cutting it in half would mean a 49.5% reliability.    But if you have two drives that are individually 99% reliable, then the pair is 0.99 x 0.99 = 0.9801, or about 98% reliable.

 

Bottom line:  I see no problem using a 2-drive RAID-0 as a parity drive if you wanted to move to SMR drives but didn't want the SMR "penalty" for your parity writes (which would improve performance if you're doing multiple simultaneous writes to the array.

 

[SSD]

I think these drives show quite a bit of promise for media applications. Will be interested to see experiences from the brave souls that try them first!

[pkn]

...

Actually, I'd be far more inclined to use RAID-0 for the parity drive than for the cache.    If one of your RAID-0 drives fails, and it's the cache drive, then everything you've written to UnRAID that hasn't yet been moved to the protected array is lost.    If it's the parity drive, then your parity "drive" just failed ... and you can recover by simply replacing the failed physical drive and doing a rebuild of the parity drive.  ...

Good point... I tend to consider cache drive as temporary place for things I (usually) have a backup for, so I'm not afraid of losing it. Now you've got me rethinking my mode of operations...

[c3]

... Note, by the way, that using a RAID-0 as parity is NOT "... cutting your drive reliability in half (or more)."    It DOES double the probability of a drive failure, but that's not the same as cutting the reliability in half.    If you have a drive that's 99% reliable, cutting it in half would mean a 49.5% reliability.    But if you have two drives that are individually 99% reliable, then the pair is 0.99 x 0.99 = 0.9801, or about 98% reliable.

 

Reliability is not a probability, but typically reported as MTBF. As you calculated, the probability is now doubled, therefore MTBF is now half.

[garycase]

Yes, the MTBF is cut in half (for any multi-drive array, it's simply (MTBF of one drive)/(Number of Drives) ... but that doesn't cut the overall reliability in half.    UnRAID (or any single parity RAID) is effectively 100% reliable unless you have a dual-drive failure ... and the likelihood of that is indeed a probability function.  I didn't bother to note it before, but the simple fact is using a 2-drive RAID-0 parity drive does NOT actually double the probability of this happening ... it does if you only consider the 2 drives in the parity array; but a more accurate calculation considers the total number of drives in the array ... so the reduction in probability is actually much less than half.  For example (using 99% as the probability of a single drive failing -- which is actually VERY low ... a more accurate number is an order of magnitude better than that for most modern drives) ... if an array has 10 drives in it, with 99% probability of failure for each drive, the likelihood of a dual drive failure is 1 - (0.99 ^ 10) = 9.56%.  Switching to a 2-drive parity array changes the drive count to 11, dropping the likelihood to 10.47% => an increase of about 10%.    Note that with more accurate probability of failure values these percentages are much lower ... e.g. at 99.9% probability the dual drive failure likelihood is 99% for 10 drives, 98.91% with 11.

 

 

[BRiT]

Reliability is not a probability, but typically reported as MTBF. As you calculated, the probability is now doubled, therefore MTBF is now half.

 

So you're saying MTBF is cut in half for every drive added to the system. So a 2 drive unRAID system has only 50% MTBF, a 3 drive system has only 25% MTBF, a 4 drive system has only 12.5% MTBF, a 11 drive system has only 0.097% MTBF, and a 24 drive system has only a 0.0000119% MTBF ? 

 

That's not how MTBF works.

[garycase]

Reliability is not a probability, but typically reported as MTBF. As you calculated, the probability is now doubled, therefore MTBF is now half.

 

So you're saying MTBF is cut in half for every drive added to the system. So a 2 drive unRAID system has only 50% MTBF, a 3 drive system has only 25% MTBF, a 4 drive system has only 12.5% MTBF, a 11 drive system has only 0.097% MTBF, and a 24 drive system has only a 0.0000119% MTBF ? 

 

That's not how MTBF works.

 

Yes, that's how MTBF is calculated -- it's simply the MTBF of the component drives divided by the # of drives.  However, your calculations aren't close to that.    For a 24-drive system, it would be 1/24th the MTBF of an individual drive ... or 0.04167 of the individual MTBF -- about 4.2%.  Not sure where you got your 0.0000119 from !!     

 

In the other cases you noted, for 3 drives it's 0.33 x MTBF, for 4 it's 0.25 x MTBF, and for 11 it would be 0.0909 MTBF => or as percentages 33% for 3 drives, 25% for 4 drives, and 9% for 11 drives.

 

Note, by the way, that in Google's study of large drive arrays they found that the probability of failure was actually higher than indicated by these numbers ... implying that MTBF is not very accurately stated by the drive manufacturers.

 

 

 

[c3]

I think he was just dividing by 2 for each drive. (MTBF hours)/(2^#drives)

 

But yes, MTBF on large scale systems means replacing drives daily (or sooner, ie hot sparing), and this shortened interval is why additional recovery methods are required (additional parity and limited stripe width). The recovery time should be less than MTBF, otherwise data loss is ensured.

 

The sole mention of MTBF in the Google report is a comment on another report, based on Seagate warranty data. The comment is that the AFR in this data set is larger than in the Seagate Waranty data set. This should not be a surprise as not all failed drives are warrantied, therefore Seagate is missing data points.

[Yousty]

Has anyone gotten one of these Seagate 8TB SMR drives yet? Really curious to find out how long preclear times are taking.

[danioj]

I've just read the entire thread and am "almost" taking a few things away from it ....

 

1. No issue with this being a parity drive (especially if the protected array is filled with these disks too);

2. No issue with this being a data drive (especially given I use a Cache Drive);

3. Read speeds are not really and issue, write speeds are (once again if your are using a Cache Drive this is somewhat moot);

4. For media storage and streaming I don't see performance downsides vs the WD Red's I am using now;

5. The cost upsides are huge. 260 x 5 (1 parity) + 150 for 500GB Cache Drive  = 32TB Storage for $1450!!!!

 

P.S. And ..... 3yr Warranty, which is what WD give their Red NAS drives!!!

 

WoW!!!!!

[garycase]

1. No issue with this being a parity drive (especially if the protected array is filled with these disks too);

 

Correct, although as you've also noted there would be a fairly significant write penalty if you aren't using a cache drive.

 

 

2. No issue with this being a data drive (especially given I use a Cache Drive);

 

Correct again -- since you're using a cache drive you've mitigated the write performance issue; and reads aren't a problem.

 

 

3. Read speeds are not really and issue, write speeds are (once again if your are using a Cache Drive this is somewhat moot);

 

Exactly.  Read speeds aren't an issue at all ... in fact, with the areal density of these drives they should be VERY good.  [Not sure what the seek times are (probably not great), but the sustained data rate will be excellent.]

 

 

4. For media storage and streaming I don't see performance downsides vs the WD Red's I am using now;

 

Agree.    The only negative relative to the Reds is the write speeds; and as you've noted a cache drive nicely mitigates this  [Who really cares what the write speeds are as long as they're all taking place when you aren't waiting for them  ]

 

 

5. The cost upsides are huge. 260 x 5 (1 parity) + 150 for 500GB Cache Drive  = 32TB Storage for $1450!!!!

 

Definitely -- at $260 that's only $32.50/TB !!    Also, a 500GB cache won't cost near $150 ... you can get a nice 500GB WD Black for $67.99 -- or a 1TB version for $74.99.

 

 

 

[WeeboTech]

From the specs, it says the MAX sustained read/write speeds are 190MB/s with average being 150MB/s.

That's not slow by today's standards. I have 1Tb drives that average from 30-90MB/s.

 

 

We'll have to wait for the reviews to trickle out before we know the truth in how slow these drives are going to be.

In any case, investing in a hardware raid controller for use with RAID0 parity or hybrid safe mode(RAID0/1), can also help alleviate the parity bottleneck if they are slow.

 

 

[MvL]

Interesting drive and a interesting thread. Which brave soul will test this drive?

[garycase]

From the specs, it says the MAX sustained read/write speeds are 190MB/s with average being 150MB/s.

That's not slow by today's standards. I have 1Tb drives that average from 30-90MB/s.

 

We'll have to wait for the reviews to trickle out before we know the truth in how slow these drives are going to be.

In any case, investing in a hardware raid controller for use with RAID0 parity or hybrid safe mode(RAID0/1), can also help alleviate the parity bottleneck if they are slow.

 

The spec sheet does say "read/write", but I doubt the writes will achieve 150MB/s except perhaps for small enough writes that the data all fits in the drive's buffer and the extra time for the rewrite of adjoining tracks is transparent.    For large (e.g. media) files, that's not likely to be the case.

 

Note that for the outer track 190MB/s speeds, the specs specifically limit the spec to reads:

"... Max Sustained Data Rate, OD Read (MB/s)  190 "

 

The specs don't mention the rotation rate, but with 1.33TB/platter areal density it must be fairly slow to be limited to 190MB/s transfers ... certainly no more than 5400rpm and possibly even slower.  Clearly they have excellent read speeds ... the real question is the write performance (which will not be nearly as good).

 

Note that these are classified as an "Archive HDD" => basically intended for write-once, read-many applications with data that doesn't change.    Whether they're a good choice for UnRAID largely depends on whether your server is used primarily for static data (movies, music, photos, etc.), or if you use it as a primary store for some of your PCs.    They're likely NOT a really good choice for the parity drive, although as I noted earlier it doesn't really matter if all of your writes are cached ... although if I was building a server with these as the data drives I'd most likely use a 2-drive RAID-0 with 4TB Reds (or Red Pros to get 7200rpm) as the parity drive.

 

 

Interesting drive and a interesting thread. Which brave soul will test this drive?

 

If I had any need for another server, I'd absolutely have no problem using these.  But I already have 3 UnRAID systems with all the storage I need for the next year or so.    The only question would be what to do for parity ... and by the time I build another system there may be an 8TB non-shingled option for the parity drive.    If not, I'd probably try an SMR drive, and if I didn't like the performance, just put in a 2-drive RAID-0 as I noted above.

 

[WeeboTech]

They're likely NOT a really good choice for the parity drive, although as I noted earlier it doesn't really matter if all of your writes are cached ... although if I was building a server with these as the data drives I'd most likely use a 2-drive RAID-0 with 4TB Reds (or Red Pros to get 7200rpm) as the parity drive.

 

 

I do not believe they will be good for parity also. I would go with the RAID0 for parity. You have to consider how often the basic functions of houskeeping come into play i.e. the journal, directories and other internal metadata.

[danioj]

http://www.amazon.com/Seagate-ST8000AS0002-ARCHIVE-128MB-3-5IN/dp/B00QX0ZGO6/ref=sr_1_3?ie=UTF8&qid=1419927800&sr=8-3&keywords=seagate+archive

[danioj]

I'll wait for the HGST's.

 

http://www.extremetech.com/computing/189813-western-digital-unveils-worlds-first-10tb-hard-drive-helium-filled-shingled-recording

 

I wonder how these will compare to the air filled Seagates and if their suitability for a nas device will equal or exceed the Seagates. WD also saying they are going to be cheaper p/TB than the Seagates too.

 

Rig full of 10TB Drives in the not so distant future. Exciting times!!

[garycase]

Actually I think the 8TB HGST helium drives are more interesting -- these use 7 platters with standard PMR recording technology, so there won't be any penalty for writes.    The 10TB units are shingled drives, so they'll have the same low write speeds that the Seagate 8TB units do.

 

In fact, although the price is likely to be pretty high (at least initially), the 8TB helium unit could be used as the parity drive for a system with the 8TB Seagate SMR drives as the data drives (assuming they're notably less expensive -- otherwise I'd just use all helium units).

 

[danioj]

Actually I think the 8TB HGST helium drives are more interesting -- these use 7 platters with standard PMR recording technology, so there won't be any penalty for writes.    The 10TB units are shingled drives, so they'll have the same low write speeds that the Seagate 8TB units do.

 

Oh, I thought these 10pTB drives were HGST drives.

 

"In what can only be described as the hard drive equivalent of Game of Thrones, Western Digital’s HGST has announced the world’s first 10-terabyte hard drive: the helium-filled Ultrastar He10."

 

I'm enthusiastic but still learning!!! Sorry!!