Multi-Actuation: Seagate working to double IOPS on Spinning Disks

[WMGroomAK]

As a part of Seagate's push for more data center drives, they are looking at adding in multi-actuator HDDs to their enterprise drive lines in 2018.  With their proposed multi-actuator drives, Seagate thinks that they can double the data performance for reads and writes as compared to an existing Hard Disk, which is one of the biggest bottlenecks for deployment of High Capacity drives.  

https://techreport.com/news/32976/seagate-multi-actuator-hard-drives-think-two-heads-are-better-than-one

Quote

Technologies like shingled magnetic recording (SMR), heat-assisted magnetic recording (HAMR), and microwave-assisted magnetic recording (MAMR) all promise to improve the capacity of spinning platter drives, but none of them were developed with performance in mind. While HAMR and SMR could indirectly offer better throughput by way of increased data density, large jumps in hard drive performance have been rare in recent years. Seagate says its multi-actuator technology could double hard drive performance while maintaining magnetic hard drives' advantages over SSDs in cost per storage unit.

Simultaneous operation of two heads could double the throughput of a drive or double the drive's maximum IOPS capability. Seagate says multi-actuator drives could present two parallel access streams to the same drive. The heads attached to either actuator can perform two reads, two writes, or one read and one write operation simultaneously.

 

Paul Alcorn of Tom's Hardware says that Seagate's multi-actuator drives have a second magnet that couples with the voice coil magnet assembly, implying that manufacturing costs will be somewhat higher than conventional drives. Management of two heads will require a more sophisticated controller and more complex firmware than current drives, but according to Alcorn, this probably represents less of a challenge. He also notes that weight and power consumption of devices with the new technology will probably increase somewhat.

 

The company says the multi-actuator technology could be extended to drives with more than two actuators in the future, given refinements to the manufacturing process. Seagate didn't provide any pricing or availability information, but the wording of the company's blog post suggests that multi-actuator drives will come to the data center first.

This is kind of interesting in that you would have each set of heads able to independently access separate partitions of the platter stack, although I'm thinking that the issuing of the changes in magnetic fields, high speed spinning of all of those platters and potential imbalance issues from two separate actuator motors probably made this an engineering nightmare...  I also severely doubt the claims of doubling the data performance, but would really like to have them achieve that.  Of course, I would also like to see an improvement on some of their quality issues as I've had several Seagate drives fail on me in the past. 

 

Seagate Blog Post: https://blog.seagate.com/technology/multi-actuator-technology-a-new-performance-breakthrough/?lipi=urn%3Ali%3Apage%3Ad_flagship3_messaging%3BJXyixoGaROyvrtBYgUlcBw%3D%3D

[GDRRiley]

well lets see how long it takes for them to implement or say that they failed.

sounds interesting way to do it. 

[Syntaxvgm]

seems like itll be loud 

 

[Dabombinable]

Wouldn't it also be possible to have every head actuate independently with current technology?

[ARikozuM]

I'd like to see them start this on consumer 2-4TB drives instead. I wouldn't mind paying $150 for a faster 2TB NAS.

[WMGroomAK]

3 minutes ago, ARikozuM said:

I'd like to see them start this on consumer 2-4TB drives instead. I wouldn't mind paying $150 for a faster 2TB NAS.

Would really like to see this on some 4 to 6 TB consumer NAS drives myself...  Would be really interesting to see how much better a parity check on my home array would be.

[ARikozuM]

1 minute ago, Dabombinable said:

Wouldn't it also be possible to have every head actuate independently with current technology?

I don't think it is, at least not now. The piezoelectric motor itself is quite small, but the main actuator is quite large. I'd be interested in seeing a schematic for how the actuators are attached with this design. 

[knightslugger]

it would still take one full revolution of the platter to access that same data. Their idea sounds a lot like RAID 0 contained within one drive.

[Curufinwe_wins]

9 minutes ago, ARikozuM said:

I don't think it is, at least not now. The piezoelectric motor itself is quite small, but the main actuator is quite large. I'd be interested in seeing a schematic for how the actuators are attached with this design. 

It's also a large decrease in reliability. Actuators are already pretty much the main physical failure mechanisms for HDD, so i'll be interested to see reliability claims when these come out.

 

7 minutes ago, knightslugger said:

it would still take one full revolution of the platter to access that same data. Their idea sounds a lot like RAID 0 contained within one drive.

Not quite. This design explicitly helps for burst applications. Properly optimized sequential loads would preform identically as a large bit of technology already does into ordering data on the platters such that data can be read as continuously as possible.

[jagdtigger]

15 minutes ago, ARikozuM said:

I'd like to see them start this on consumer 2-4TB drives instead. I wouldn't mind paying $150 for a faster 2TB NAS.

You can do it right now, its called RAID0... This thing is going to be similar, the only difference is that its done inside the drive.

[Matu20]

Looks like history.

[ScratchCat]

Finally.

Spoiler

At a random guess wouldn't it be possible to add another actuator in the empty space on the right, further away from the magnetic fields of the other actuator? Although bringing another set of strong magnets near the RW heads would be a bad idea. The head parking position could be dealt with by sectioning each layer so they can pass through each other safely.

[potoooooooo]

Soo...double the chance of failure. Got it.

[ARikozuM]

7 minutes ago, jagdtigger said:

You can do it right now, its called RAID0... This thing is going to be similar, the only difference is that its done inside the drive.

We're moving the posts with this metaphor. This keeps things in one slot rather than splitting the difference between two or more. 

[Rune]

Seems loud, hot, and shaky as hell.

[WMGroomAK]

5 minutes ago, ARikozuM said:

We're moving the posts with this metaphor. This keeps things in one slot rather than splitting the difference between two or more. 

But going with this metaphor, you could end up with a RAID 0 inception.  I mean if it is an internalized RAID 0 on a single HDD and you set multiple up as RAID 0, does that make it a RAID 00? 

[jagdtigger]

2 minutes ago, WMGroomAK said:

But going with this metaphor, you could end up with a RAID 0 inception.  I mean if it is an internalized RAID 0 on a single HDD and you set multiple up as RAID 0, does that make it a RAID 00? 

Putting aside the funny part (00 = toilette ) i think yes, but IDK how much that would mean in real world performance...

[ARikozuM]

6 minutes ago, WMGroomAK said:

But going with this metaphor, you could end up with a RAID 0 inception.  I mean if it is an internalized RAID 0 on a single HDD and you set multiple up as RAID 0, does that make it a RAID 00? 

Okay, SSD's run on RAID 0 now...

[Guest]

2 hours ago, WMGroomAK said:

each set of heads able to independently access separate partitions of the platter stack

Man I would love a few of these for VM storage. With SSDs still costing a premium for capacity (Although shrinking) and my experiences with HDDs "locking up" at 100%, this seems like a great boon for system responsiveness. The only question I have is how much would designing around two actuators increase the costs.

1 hour ago, potoooooooo said:

Soo...double the chance of failure. Got it.

While it does heighten the chance of a single actuator failing in the system, I would think approach would be more reliable as long as they designed it to properly function with only one working in case of failure.

Edit: Nvm, maybe I should have looked at the picture. For some reason I was under the impression that the heads would be on opposite sides of the drive, not stacked on each other.

Edited December 20, 2017 by Guest

[TheRandomness]

Thing is, with our technology, the main question is why hasn't this been done already?

[XenosTech]

15 minutes ago, TheRandomness said:

Thing is, with our technology, the main question is why hasn't this been done already?

Because fuck innovation

[Dabombinable]

35 minutes ago, TheRandomness said:

Thing is, with our technology, the main question is why hasn't this been done already?

I've been thinking for years about something like this-except across all platters (preferably independently on both sides of each.

20 minutes ago, XenosTech said:

Because fuck innovation

2010 2.5" 7200RPM 500GB HDD. Manages to do 120-140MB sec sustained. Still not seeing that much of an improvement 7 years later with current 7200RPM 2.5" HDD.

[Drak3]

37 minutes ago, TheRandomness said:

Thing is, with our technology, the main question is why hasn't this been done already?

Probably a shit ton of validation as to avoid pushing a problematic product.

[XenosTech]

1 minute ago, Dabombinable said:

I've been thinking for years about something like this-except across all platters (preferably independently on both sides of each.

2010 2.5" 7200RPM 500GB HDD. Manages to do 120-140MB sec sustained. Still not seeing that much of an improvement 7 years later with current 7200RPM 2.5" HDD.

I'm honestly surprised they didn't make 10k drives cheaper for consumers by now too

[ARikozuM]

42 minutes ago, TheRandomness said:

Thing is, with our technology, the main question is why hasn't this been done already?

Motors would take up a lot of space, data and voltage delivery would have to be rethought, and heat is a detriment that has finally been pushed due to different gases being tested over conventional air.