News Western Digital Ships 22TB HDDs for Mass Market

[Admin]

Western Digital begins to sell 22TB WD Gold, WD Red Pro, and WD Purple Pro hard drives.

Western Digital Ships 22TB HDDs for Mass Market : Read more

 

[brandonjclark]

For those interested, over a SATA 3 port this would take nearly a day to copy off 22TB.

 

[Kamen Rider Blade]

This is why "Multi-Actuator" technology with Internal RAID-0 across the seperate platters needs to become standard to keep up with the R/W speeds to make copying 22 TB reasonable.

Lose 1 platter, but add one more actuator Stack and you could go from 291 MB/s -> 582 MB/s.
9 Platter, 10 Actuator Arms -> 2x Actuator Stacks of 5-Arms a piece.

That comes awfully close to maxing out SATA 3.0 - 600 MB/s bandwidth cap.

Maybe they should give us SATA 4.0 - 1200 MB/s bandwidth cap and let us have more actuators.

Imagine what 3x & 4x stacks of Actuators could do.

That much more R/W throughput.

With a little bit more R&D:

Eventually getting it down to each arm being it's own Stack where every Actuator Arm is fully independent!
Imagine a world where 10 Platters, 11 Actuators, all fully independent & R/W in RAID 0.
11*291 MB/s = 3201 MB/s.
SAS-4 = 2.8125 GB/s, that's not enough.
SAS-5 = 5.625 GB/s, that's enough bandwidth.

The eventual technological road you can go down with using Multi-Actuator technology easily allows you to make R/W of 1 TB reasonable when you can easily pump up R/W speeds into the GB range.

 

[BX4096]

Lose 1 platter, but add one more actuator Stack and you could go from 291 MB/s -> 582 MB/s. ... Imagine a world where 10 Platters, 11 Actuators, all fully independent & R/W in RAID 0...

While it sounds like an awesome no-brainer to a layman, I would assume that there are good reasons (aside from the added cost, of course) why it hasn't been implemented yet. If anyone know what they are, I'd like to hear more about it.

Frankly, I got surprised when I learned of SATA's undignified demise. I still use plenty of high-capacity HDDs on my machines, and I doubt I'll see a reason to abandon them altogether in my lifetime. If they could upgrade this "dying" technology this way, I'd be ecstatic.

 

[Geef]

The reason: Cost of failure.

Adding an extra actuator stack means that many more pieces that could fail. Adding a single platter instead lowers the number of total pieces and chance of failure.

 

[USAFRet]

For those interested, over a SATA 3 port this would take nearly a day to copy off 22TB.

That is when you fail over to your backup, and let the "recovery" from this 22TB drive take its own sweet time.

 

[Grobe]

The reason: Cost of failure.

Adding an extra actuator stack means that many more pieces that could fail. Adding a single platter instead lowers the number of total pieces and chance of failure.

I also wonder if failure of one R/W head, then all content of the hdd are lost with that kind of setup.

 

[USAFRet]

I also wonder if failure of one R/W head, then all content of the hdd are lost with that kind of setup.

1 head, multiple heads...this is what backups are for.

If you can't manage backups, you don't need a 22TB drive.

 

[Grobe]

If you can't manage backups, you don't need a 22TB drive.

Of course - cannot be said enough, but another discussion in my view.

I should have put my words different: What I'm curious about is if many platters like this affects the expected lifetime of the product, and if different methods of internal distributing of data on the platters also affect MTBF.

 

[spongiemaster]

I also wonder if failure of one R/W head, then all content of the hdd are lost with that kind of setup.

The only currently sold multi actuator hard drive, Seagate's MACH.2, requires both actuators to be working properly for the drive to operate.

 

[spongiemaster]

Lose 1 platter, but add one more actuator Stack and you could go from 291 MB/s -> 582 MB/s.
9 Platter, 10 Actuator Arms -> 2x Actuator Stacks of 5-Arms a piece.

Would assume that you would lose a platter for every additional actuator. So in a standard 3.5 HD case, you'd only get 5 platters with their own actuator, losing over half the potential capacity in the process. That would result in an insanely expensive per GB mechanical hard drive not to mention the ridiculous complexity of such a drive and trying to tune the firmware. SSD's would be more cost effective at that point.