2021 edition: best HDDs for archiving

[JHZR2]

Every year or so, as I get some new tech or other reason to think about it, I have some question and/or some plan about archiving.

Here we are almost to 2021, and I’m thinking about archiving again, especially the “keep three copies” aspect. I know we have discussed data longevity on ssd drives. I’ve also read that optimal archiving on HDDs requires a rewrite every three years to re-magnetize.

I’m going to get a new HDD, and maybe a thunderbolt and/or USB 3.2 gen 2 type interface.

So, for preparing for 2021, a few questions:

1) who is currently making the most reliable platter type HDDs in the 2.5 and 3.5” form factor?

2) which form factor is longer lived? More robust? I’d suspect that the smaller form factor drives have benefits due to being made for mobile applications and having smaller parts that can’t vibrate/move as far.

3) Is there a benefit or disadvantage to buying a premade usb HDD from one of the major brands, versus bare drives and enclosures or a SATA-USB connector?

Thanks!

 

[Skippy722]

I’d go with a 3.5” HDD personally in a SATA-USB enclosure. This seems a bit overkill, but has a claimed MTBF of up to 2.5 million hours.

WD Gold Enterprise Class SATA Hard Drive | Western Digital Store

Conquer tough workloads with enterprise-class WD Gold™ HDDs. Available in 1TB to 18TB1 capacities, this highly reliable solution for demanding storage environments provides up to 2.5M hours MTBF, vibration protection technology and a low power draw thanks to HelioSeal™ technology [for 12TB and...

shop.westerndigital.com

 

[uc50ic4more]

2) which form factor is longer lived? More robust? I’d suspect that the smaller form factor drives have benefits due to being made for mobile applications and having smaller parts that can’t vibrate/move as far.

That's a good point; and it would be interesting to see some data or hear some anecdotal observations about the inevitability of "turbulence" in the life of a 2.5" (mobile USB cases, laptops, etc.) and their lifespan versus an off-the-shelf 3.5" drive. However, I know that it isn't hard or expensive to get "enterprise"-grade 3.5" meant for the rigours of life in a server whose "Mean Time Between Failure" can be very impressive.

 

[JHZR2]

That's a good point; and it would be interesting to see some data or hear some anecdotal observations about the inevitability of "turbulence" in the life of a 2.5" (mobile USB cases, laptops, etc.) and their lifespan versus an off-the-shelf 3.5" drive. However, I know that it isn't hard or expensive to get "enterprise"-grade 3.5" meant for the rigours of life in a server whose "Mean Time Between Failure" can be very impressive.

I have also recently read that they make enterprise grade 2.5" drives, because they are less clunky, more can be fit, and because the speed can be better since tip speed is limited on the larger disc.

Havent looked hard enough to find enterprise grade 2.5" drives though...

 

[PandaBear]

I have also recently read that they make enterprise grade 2.5" drives, because they are less clunky, more can be fit, and because the speed can be better since tip speed is limited on the larger disc.

Havent looked hard enough to find enterprise grade 2.5" drives though...

The calculation from what I heard, is that you simply cannot get data into / out of a drive fast enough with these massive drives on the same connection speed. In a way I think it is sad that a computer is tied down to the drive because they don't support too many connection, when they could be used to do constant calculations. I can't wait till the day we can switch hard drives and SSD to ethernet connection so we don't have to waste processor bandwidth on.

 

[JHZR2]

The calculation from what I heard, is that you simply cannot get data into / out of a drive fast enough with these massive drives on the same connection speed. In a way I think it is sad that a computer is tied down to the drive because they don't support too many connection, when they could be used to do constant calculations. I can't wait till the day we can switch hard drives and SSD to ethernet connection so we don't have to waste processor bandwidth on.

What am I missing WRT that? One can barely get 1GBPS internet/Ethernet connections outside of very specific LANs. SATA does what? 6GBPS? thunderbolt does 20?

 

[PandaBear]

What am I missing WRT that? One can barely get 1GBPS internet/Ethernet connections outside of very specific LANs. SATA does what? 6GBPS? thunderbolt does 20?

Enterprise is not 1gbps, more like starting at 25-40 and ending at 100gbps.

SATA / SAS won't be fast enough to serve the amount of data stored inside 1 drive (i.e. 13-16TB), so basically they will need to split up the drive's payload to do near line and long term storage to justify this capacity.

 

[JHZR2]

Enterprise is not 1gbps, more like starting at 25-40 and ending at 100gbps.

SATA / SAS won't be fast enough to serve the amount of data stored inside 1 drive (i.e. 13-16TB), so basically they will need to split up the drive's payload to do near line and long term storage to justify this capacity.

So you’re saying that when 100GB type connections become ubiquitous in the future, no need for local storage because retrieval from “the cloud” will be so much faster than anything local? Wouldn’t it still be limited by how fast the multiplexed raid arrays or whatever the future brings can source those data?

 

[PandaBear]

So you’re saying that when 100GB type connections become ubiquitous in the future, no need for local storage because retrieval from “the cloud” will be so much faster than anything local? Wouldn’t it still be limited by how fast the multiplexed raid arrays or whatever the future brings can source those data?

I am not saying that. What I am saying is, within a large datacenter, their HDD is being used all day non stop, and their array is bottlenecked when HDD is getting too large.

The array would still be the bottleneck, but if the array is cheaper (say, without needing a Xeon only to shuttle data between fiber / 40gbps ethernet to a SAS / SATA drive), they can use more arrays and make the bandwidth problem less of a problem by slowing down the amount of data going in and out of each TB of capacity.

 

[JHZR2]

Thanks for the explanation!