[HN Gopher] Seagate: 100TB HDDs Due in 2030, Multi-Actuator Driv...
___________________________________________________________________
Seagate: 100TB HDDs Due in 2030, Multi-Actuator Drives to Become
Common
Author : ItsTotallyOn
Score : 15 points
Date : 2021-03-06 20:58 UTC (2 hours ago)
(HTM) web link (www.tomshardware.com)
(TXT) w3m dump (www.tomshardware.com)
| baal80spam wrote:
| > Thank you for visiting this site. Unfortunately we have
| detected that you might be running custom adblocking scripts or
| installations that might interfere with the running of the site.
|
| This is a sure way to make me not visit a site ever again.
| nottorp wrote:
| Works fine with firefox/ublock origin, what else do you use?
|
| But speaking of idiocies like that, rockpapershotgun won't let
| me see subscription prices unless i agree to tracking cookies.
| Do they really think that's an incentive to subscribe?
| Exmoor wrote:
| I'd appreciate more capacity, but at this point my big complaint
| is price per TB. I've been paying $15/TB for drives for, I think,
| 3-4 years. I'm beginning to wonder if we've hit the end of the
| line for decreasing storage costs.
| magicalhippo wrote:
| Speed is also getting to be an issue. Replacing a 14TB disk in
| a mirror can take almost a day (mine do ~220MB/s average).
|
| SATA is maximum of 6Gbps, but even maxing that (750MB/s) for a
| 100TB drive it would still take 37 hours to mirror a full disk.
| bcrl wrote:
| Meanwhile, 100TB SSDs already exist and can be purchased today.
| They're incredibly expensive at $40,000, but if you really need
| one... See https://nimbusdata.com/products/exadrive/pricing/ for
| details.
| nottorp wrote:
| I forget... is HAMR the slow one that kills your NAS?
| liminalsunset wrote:
| I think you're thinking about SMR - Shingled Magnetic
| Recording. The tracks overlap because the heads are larger for
| writing than for reading, so writing can be a destructive
| process and data must be rewritten when adjacent data is
| written.
|
| These types of drives are fast for sequential writes, but any
| operations that edit data may be slow. Additionally, the drive
| spends time when idle clicking around to "defragment" the
| shingles IIRC.
| bcrl wrote:
| SMR drives are not "fast" for writes. They run at slightly
| less than the same speed as non-SMR drives do for sequential
| writes (opening zones has a cost). They are unpredictable and
| often slow for everything else. SMR drives are somewhat like
| the first generation of SSD controllers: they lots of glass
| jaws that nobody ever expects when garbage collection begins
| to kick in.
| tinus_hn wrote:
| > they lots of glass jaws that nobody ever expects
|
| Looks like autocorrect got involved here
| magicalhippo wrote:
| As noted that's SMR[1], which is more or less orthogonal to
| HAMR[2]. That is, AFAIK, a drive can have HAMR without using
| SMR, or it can use both.
|
| [1]: https://en.wikipedia.org/wiki/Shingled_magnetic_recording
|
| [2]: https://en.wikipedia.org/wiki/Heat-
| assisted_magnetic_recordi...
| nottorp wrote:
| Ah thanks. The pessimist in me is expecting all new models to
| have both though :)
| nabla9 wrote:
| 4 Gb/mm2 for HDD density is a lot.
|
| For comparison:
|
| 7 Gb/mm2 - current NAND technology
|
| 0.3 Gb/mm2 - current DRAM technology
|
| 0.2 GTr/mm2 - current FinFET technology
| dogma1138 wrote:
| Pure density isn't everything, magnetic storage doesn't have
| the same power and heat increases with storage density as solid
| state semiconductor storage has.
___________________________________________________________________
(page generated 2021-03-06 23:02 UTC)