60TB Hard Drives within the decade

[garycase]

... 6TB drives are more expensive per TB vs 4TB, and 4TB are more expensive than 3TB drives ...

 

These differences are, however, very nominal => not enough to matter in the great scheme of things.    Using one drive type (WD Red) and one retailer (Newegg), for example, today's prices are $64.99/TB for 1TB;  $46.50/TB for 2 TB;  $39.67/TB for 3TB;  $40.00/TB for 4TB;  and $41.67/TB for 6TB units.    Note that the 3TB, 4TB, and 6TB units are effectively the same cost/TB.    Other drive makes/models will change this relationship a bit; but not by dramatic amounts.

 

And when you factor in the support costs per drive [sATA ports, power, cooling, drive slots, etc.] it makes even more sense to use larger drives, especially if the /TB cost is only nominally different.

 

WD doesn't make an 8TB unit, but for comparison purposes, the 8TB Seagate Archive unit is currently $32.50/TB at Newegg.

 

I don't think it's a stretch at all to expect a 60TB drive to be under $1000 ... that's only half of the /TB cost of the 8TB archive units today.  Could even be notably lower.

 

... and as C3 noted, a 60TB drive will almost certainly be more like an archive unit than a high-speed workstation drive.    For high speed boot devices SSDs will rule (they pretty much already do).

 

... I also agree with C3 on one simple fact:  "... $/TB for all storage will continue to drop because without a lower $/TB the drives wont be made."

 

[c3]

$/TB is usually something quoted for drives themselves. Otherwise it is not something we can objectively compare.

 

Actually, the comparison is very easy and done very often.

 

The infrastructure to support X many drives is $X, regardless if the drive size. Take 6TB vs 8TB. The process is just ($X+$6TB)/(6TByield) vs ($X+$8TB)/(8TByield). As the number and size of drives grows, $X becomes less a factor. Thus the value approaches the drive $/TB. Unfortunately, this is so routine I have spreadsheet models for the various pricing and yield options. I've even attached an unRAID version to this forum on some older post.

 

But I agree - each organization has to weigh all these types of variables for their unique operations to make good business decisions. As stated, 60 TB drives would offer a lot of cost saving advantages to organizations that go through drives like water, even if the drives themselves are at a premium.

We destroy more drives per year than BackBlaze owns.

[SSD]

... 6TB drives are more expensive per TB vs 4TB, and 4TB are more expensive than 3TB drives ...

 

These differences are, however, very nominal => not enough to matter in the great scheme of things.    Using one drive type (WD Red) and one retailer (Newegg), for example, today's prices are $64.99/TB for 1TB;  $46.50/TB for 2 TB;  $39.67/TB for 3TB;  $40.00/TB for 4TB;  and $41.67/TB for 6TB units.    Note that the 3TB, 4TB, and 6TB units are effectively the same cost/TB.    Other drive makes/models will change this relationship a bit; but not by dramatic amounts.

 

And when you factor in the support costs per drive [sATA ports, power, cooling, drive slots, etc.] it makes even more sense to use larger drives, especially if the /TB cost is only nominally different.

 

WD doesn't make an 8TB unit, but for comparison purposes, the 8TB Seagate Archive unit is currently $32.50/TB at Newegg.

 

I don't think it's a stretch at all to expect a 60TB drive to be under $1000 ... that's only half of the /TB cost of the 8TB archive units today.  Could even be notably lower.

 

... and as C3 noted, a 60TB drive will almost certainly be more like an archive unit than a high-speed workstation drive.    For high speed boot devices SSDs will rule (they pretty much already do).

 

... I also agree with C3 on one simple fact:  "... $/TB for all storage will continue to drop because without a lower $/TB the drives wont be made."

 

60TB drive at $1000 = $16.67/TB. At $40/T, the cost would be $2400. Why do you think WD is going to sell these drives so inexpensively? Wishful thinking?  I expect the opposite. More like $50/T.

 

As c3 and I agreed, the $/TB of the drive is just one part of the total cost for a user of drives. For example, BackBlaze might be building only 10% the number of pods with these drives - and the pods themselves are not cheap. Plus the square footage for all of those pods, the support costs, power, etc. So in the larger sense, the price could come down dramatically even if the drive prices are higher.

[garycase]

60TB drive at $1000 = $16.67/TB

 

Yep.  And the 8TB archive drives are currently $32.50/TB.  As I noted earlier, a sub $1000 60TB drive only requires the /TB price be dropped in half from that.    In fact, at $16.25/TB (half of $32.50) the cost would be $975.    That's a VERY reasonable expectation ... in fact, it's likely to be a good bit lower.

 

 

At $40/T, the cost would be $2400. Why do you think WD is going to sell these drives so inexpensively? Wishful thinking?  I expect the opposite. More like $50/T.

 

There is NO chance that a 60TB drive will be priced at that level.  Even if it was priced at the same level as today's 8TB archive units, it would only be $1950.    And it simply won't be that expensive.    As C3 noted earlier, "...  $/TB for all storage will continue to drop because without a lower $/TB the drives wont be made."

 

I'd be glad to sell you as many 60TB drives as you want, for delivery within 6 months of availability,  for $1500 each, pre-paid NOW.  [and I'd make a nice profit on that deal  :) ]    If they're not available within 48 months I'll simply refund your $$

 

 

 

 

[SSD]

60TB drive at $1000 = $16.67/TB

 

Yep.  And the 8TB archive drives are currently $32.50/TB.  As I noted earlier, a sub $1000 60TB drive only requires the /TB price be dropped in half from that.    In fact, at $16.25/TB (half of $32.50) the cost would be $975.    That's a VERY reasonable expectation ... in fact, it's likely to be a good bit lower.

 

 

At $40/T, the cost would be $2400. Why do you think WD is going to sell these drives so inexpensively? Wishful thinking?  I expect the opposite. More like $50/T.

 

There is NO chance that a 60TB drive will be priced at that level.  Even if it was priced at the same level as today's 8TB archive units, it would only be $1950.    And it simply won't be that expensive.    As C3 noted earlier, "...  $/TB for all storage will continue to drop because without a lower $/TB the drives wont be made."

 

I'd be glad to sell you as many 60TB drives as you want, for delivery within 6 months of availability,  for $1500 each, pre-paid NOW.  [and I'd make a nice profit on that deal  :) ]    If they're not available within 48 months I'll simply refund your $$

 

ROFL! I don't think I'll be sending you thousands of dollars for safe keeping.

 

I thought you were certain that the drives would be coming out at under $1000? Suddenly you want me to pay $1500. $25/T is a deal but far from a steal of $16/T.

 

And the cost of parity of $1500 is far higher than I would pay. I would only consider 60T parity appropriate if I was looking for an array of 5-10x that size, which is not in the cards for me. Think of the parity check times - would take a week!

 

I will agree to a gentlemanly wager though. Once the facts unfold (if they do), the winner will be clear. And I will respectfully yield to your superior prognostication skills should the drive emerge for under $1000, and you to mine if the $/TB of the 60T is over the prevailing price per TB at that time.

[c3]

I will agree to a gentlemanly wager though. Once the facts unfold (if they do), the winner will be clear. And I will respectfully yield to your superior prognostication skills should the drive emerge for under $1000, and you to mine if the $/TB of the 60T is over the prevailing price per TB at that time.

 

Too much overlap in those. New drives always launch a bit high and settle down. If the 60T launches as enterprise, what are the comparables? As the listed WD Reds show, the 3TB is the ugly duck and lowest price. The new drive should not be compared to the ones just taking up shelf space.

 

Say the 60T launches for $900 and the 30T is going for $400?

[garycase]

I thought you were certain that the drives would be coming out at under $1000?

 

I am.  That's why I'd be glad to accept any risk of being wrong for a nice profit ... I'm convinced I'd make well over $500/drive if you were foolish enough to actually pay $1500 for them.    My upside risk would be FAR better than any downside risk.    If your expectation of close to $50/TB [" I expect the opposite. More like $50/T"] were to prove even close to accurate, I'd lose $1500/drive !!

 

 

I will agree to a gentlemanly wager though. Once the facts unfold (if they do), the winner will be clear. And I will respectfully yield to your superior prognostication skills should the drive emerge for under $1000, and you to mine if the $/TB of the 60T is over the prevailing price per TB at that time.

 

I didn't say the cost/TB would be under the prevailing price/TB => I said the drives will cost under $1000.  I have no idea what the cost/TB of other drives are likely to be at that time, but I'd expect them to also be much lower than today.  The cost/TB of a 60TB drive may in fact be higher than that of other drives -- my prediction is simply that it will be low enough that the drives will cost under $1,000.    (I actually think it'll be even lower)

 

 

 

 

[RNO]

We will need a dramatic increase in speed to use these babies.

We can't use them now, we could experience crash during the process of recovering or just preclearing the replacement drive.

Preclearing a 6TB drive with 3 cycles takes around 130h, imaging the time needed for a HDD 10x larger !

You can't expect that nothing would happen during this laptime.

 

And we did not talk about the following process of puting it in the array, this would take a while too.

 

We will need SAS 5.0 or SATA-5/6 level of speed to use this kind of HDD capacities.

 

just my two cents

[dirtysanchez]

We will need SAS 5.0 or SATA-5/6 level of speed to use this kind of HDD capacities.

 

Even the fastest current gen platter-based spinners top out at around 200MB/s on the outer (fastest) tracks.  The existing SATA III interface is capable of 600MB/s, fully 3 times faster than even the fastest current spinners can achieve.  So the drives themselves have a long way to go before they can even max out the existing interface.  Of course the much higher areal density of the mythical 60TB drive would likely have transfer speeds in excess of SATA III, so a faster interface would be required to take advantage of the speed, but the drive itself is far behind what the interface is capable of.

[garycase]

Clearly, as dirtysanchez noted, the areal density of drives with capacities in that range will be appreciably higher than those of today, so the sustained transfer rates will be much higher.    And the interface capabilities will almost certainly be improved as well.    Indeed, you can already get 1.2GB/s SAS controllers, and I'm sure SATA will hit those speeds soon [well before 60TB drives are available  ].

 

As an example of potential timing ... consider that a 30TB drive is only 5 times today's 6TB drives.  If this drive had 5 times the effective areal density of current 6TB drives, and a "SATA-1200" interface, then a parity check using these drives would take essentially the same time as it does today with the 6TB drives.

 

[fishface]

I wouldn't be surprised if SSD doesn't have a big impact on where spinning disk end up. Samsung have 3-4TB on the horizon, and 3D NAND will help decrease cost.  Rumour has it that Intel/Micron are planning on upsetting the SSD market in 2017.

 

Flash storage is now going in SAS tray for several new Dell servers and storage.

 

Interesting read here: http://www.theregister.co.uk/2015/06/25/pure_storages_mass_disk_drive_killer/

 

I wouldn't be surprised if the SATA interface dies out as well, replaced by M.2.