40 and 50 TB drives from Seagate

I love the idea of running fewer disks for the same capacity, because you get the power savings. In some places like Europe where energy prices are spiking, this is a much bigger deal than it used to be. But I'm not sure storage controller speed (or drive cache) has kept up with these capacity improvements.

For example, if I had 8x16TB drives in RAID6 I'd get the capacity of 6 of them (6*16=96). To get similar capacity/fault tolerance out of 32TB drives I'd need 5 (3*32=96, 2 for parity). In theory that'll yield about a three disk energy savings. But which array would rebuild faster?
1686594945520.png
 
philb2 said:
Maybe the existence of these drives will lead to a much faster SATA 4 standard
Click to expand...
HDD are not that close to saturate SATA 3 speed limit 600 MB/s for sustained long job no ?

Drive speed is the limitation more than the port, wouldn't SAS have the same issue ? I could be all wrong here
 
LukeTbk said:
HDD are not that close to saturate SATA 3 speed limit 600 MB/s for sustained long job no ?

Drive speed is the limitation more than the port, wouldn't SAS have the same issue ? I could be all wrong here
Click to expand...
LukeTbk

OK, I should have written out my assumption that Seagate (and WD?) will be able to improve transfer rate due to increased bit density.

The initial customers for these drives will be data centers and cloud services. They will tell Seagate what they need to make these drives practical.

Pure speculation, for shits and grins. Imagine if Seagate or WD used this technology in a 5 1/4" form factor. Or to really go back in time, 8" form factor. https://www.storagenewsletter.com/2018/07/22/history-1979-shugart-associates-sa1000/
 
philb2 said:
OK, I should have written out my assumption that Seagate (and WD?) will be able to improve transfer rate due to increased bit density.
Click to expand...
They could double it and still be ok, latest dense HDD reach 275 mb/s which is significantly faster than in the past but still a lot of room from 600MB/s.

Old 15,000 RPM ultra fast SCSI HDD in the past reached 160mb/s for example
 
philb2 said:
Imagine if Seagate or WD used this technology in a 5 1/4" form factor.
Click to expand...

Market failure cause nobody has 5.25" bays... But I once got some surplus 5.25" full height drives, model Seagate Elite 47. They were monsters, with 28 heads and weighed 7 pounds each. Of course, the drives speced here have 10 platters and 20 heads in a regular 3.5" form factor. I'm guessing you could get at least 40 platters in a full height 5.25", and the platters would be at least double the area, so like 8x the storage. Likely doesn't make sense though.
 
philb2 said:
Red Falcon

Maybe the existence of these drives will lead to a much faster SATA 4 standard. Or maybe SAS ports on consumer motherboards. Or maybe neither. :wtf:
Click to expand...
I doubt it, the SATA protocol has been maxed out for over a half-decade now, and HDDs don't really come close to saturating it even in sequential transfers.
Even the dual-actuator HDDs, as mentioned above, only get up so high, and an array rebuild just takes time to run through the whole platter, which on something upwards of 50TB would probably take over a week to rebuild, especially so with a single-actuator HDD.

Other than cold storage or archiving data, these drives are almost obsolete out of the gate just due to their sheer size and minuscule transfer rates.
 
toast0 said:
I'm guessing you could get at least 40 platters in a full height 5.25", and the platters would be at least double the area, so like 8x the storage. Likely doesn't make sense though.
Click to expand...
Rotational speed would take a hit as a consequence without new and amazing platter tech and the arms would also increase in size. Latency/seek times would massively increase.

10k/15k rpm 3.5" drives were basically tall 2.5" drives on the inside.
 
Still want the sweet spot for price/performance and capacity. Spinning disk is used for large media storage only. Don't care about performance much.
 
LukeTbk said:
They could double it and still be ok, latest dense HDD reach 275 mb/s which is significantly faster than in the past but still a lot of room from 600MB/s.

Old 15,000 RPM ultra fast SCSI HDD in the past reached 160mb/s for example
Click to expand...

If density increases equally in both dimensions, then it would take a 4x (per platter) density increase to double the transfer rate.
 
Red Falcon said:
Other than cold storage or archiving data, these drives are almost obsolete out of the gate just due to their sheer size and minuscule transfer rates.
Click to expand...
So why did Seagate release these drives? Presumably they did some market or customer research.

And the larger question is: what is the future of spinning HDDs? How many years until SSDs dominate the market for "normal" use, even say 8 ot 16 TB drives for consumer use. Personally I could certainly use an 8 TB SSD to store photos (Nikon NEF), music, books, etc. But the price of an 8 TB NVMe drive is way more than I can justify.
 
Not sure about the minuscule transfer rate, if you have say only 6 of them in a fast ZFS mode, it would already saturate 10gb no ? And 16 drive pool would need 40+gb type connection, well maybe that count as minuscule transfer rates now a day.

I imagine it is not uncommon for HDD storage pool to still have more possible theoretical max bandwidth than their network connexion allow, those drive are very slow for small file, but for large file transfer 250-275mbs by second is not that far from sata SSD (specially over long sustained work).
 
Last edited:
philb2 said:
So why did Seagate release these drives? Presumably they did some market or customer research.
Click to expand...
You are giving these largecorps far more credit for intelligence and "customer research" than they deserve.

LukeTbk said:
Not sure about the minuscule transfer rate, if you have say only 6 of them in a fast ZFS mode, it would already saturate 10gb no ? And 16 drive pool would need 40+gb type connection, well maybe that count as minuscule transfer rates now a day.

I imagine it is not uncommon for HDD storage pool to still have more possible theoretical max bandwidth than their network connexion allow, those drive are very slow for small file, but for large file transfer 250-275mbs by second is not that far from sata SSD (specially over long sustained work).
Click to expand...
Do ZFS arrays scale out linearly with drives?
Traditional RAID, even RAID 0, starts to bottom out on sequential transfer rates after about 6-8 HDDs with anything added afterwards having negligible transfer rate improvements.
 
I'm also in the camp that doesn't need much performance, and 1 week rebuilds aren't a problem (although I find HDDs pretty reliable, I have 19 14TB WD taken out of external drives and no failure after 3 years), with RAIDZ3 providing great safety with one drive down.
If performance is needed SSDs are king, and are now available at reasonable prices up to a few TBs so should be enough for plenty of uses like running a massive database. At work on virtualized servers we're definitely going all SSD including keeping existing servers but replacing storage arrays from spinning rust to flash.

For years I got used to getting cheaper and bigger drives all the time, then there were the Thailand floods and things never got back to how they were before.

Early 2020 during lockdown I got these 14TB NAS drives for 250€/each, now more than 3 years later 250€ still gets me 14TB drives, maybe 16TB if I look hard, or 18TB if I add a few euros, but nothing exciting.
 
You must log in or register to reply here.
Back
Top