(edit: CES) Two arms are better than one - Seagate release dual actuator HDD for Microsoft

[Mira Yurizaki]

On the notion of "internally RAID0" and adding more things decreases fault tolerance, here's some thoughts:

• Multi-platter drives by default are "internally RAID0." And most HDDs are multi-platter
• Having two independent actuators may not decrease fault tolerance. It's not confirmed the two sets of heads will actually run in a RAID0 style, but may run in JBOD or something similar. In any case, having two independent actuators means if one is toast, the drive is still useful, just that half the platters are not. So depending on how the data is laid out when one group dies, some data my not be lost and you can still repurpose the drive for other uses until a replacement can be made.

[rcmaehl]

Single head per platter + platter based data parity when?

Could easily recovery from a single headcrash/loss of a platter with just a scratch disk and rebuilding from parity. 

[elfensky]

6 hours ago, porina said:

I struggled to see much advantage to this when it was first shown, and I still don't get it now. It's essentially two hard disks in one box, with the advantages and disadvantages that comes with it. There's still one head per side, so the only advantage is that the heads on each half of the drive can move independently of each other? I think having more than one head per side would help more, both in increasing throughput per platter and seek times.

I suppose that instead of being one slow 14tb drive it's two slow 7tb drives within 1 slot (so depending on how the system sees the drive/how the drive manages data, it could have some raid-like stripe benefits) but I agree that more heads per drive seems to be a better solution no1 is seemingly trying.

[Donut417]

8 hours ago, williamcll said:

if it isn't as cheap as QLC/PLC SSDs I doubt it would be competitive.

Hard Drives still have something over SSD's. Capacity and price. 

 

On newegg a 10 TB hard disk is $250~ just over $300 USD. I cant even find a 10 TB SSD on newegg. The biggest I found was a Samsung 7.6TB for $1,499. I can litterally buy like 4 10TB hard disks for that price. The price isnt going to be a concern for someone who is looking for mass storgate and needs a bit of performance. 

[dgsddfgdfhgs]

double failure rate?

[Anomnomnomaly]

8 hours ago, RejZoR said:

I wonder if HDD's could still be viable solution for the future if we manage to split heads to individual ones between each platter. Imagine having 1 head per platter, so you'd have 4 independent heads on 4 platter drives. It could read and write to 4 platters fully independently.

Now double it, being able to read the top and bottom side of each platter independently. 

[Bombastinator]

I don’t know how this tech would do much for consumer drives.  They’re mostly single platter. Especially for laptops.  I vaguely recall some company or other in the past actually putting more than one drive arm opposite each other (so 2 read/write heads a side) on a single platter in an attempt to increase speed.  I don’t remember how it turned out.

[Mihle]

13 minutes ago, Bombastinator said:

I don’t know how this tech would do much for consumer drives.  They’re mostly single platter. Especially for laptops.  I vaguely recall some company or other in the past actually putting more than one drive arm opposite each other (so 2 read/write heads a side) on a single platter in an attempt to increase speed.  I don’t remember how it turned out.

Any HDD with 4TB have more than one I think.

[Bombastinator]

4 minutes ago, Mihle said:

Any HDD with 4TB have more than one I think.

true But are they the top selling ones?  I just did a pcpartpicker search on mechanical hard drives by popularity.  The top 4 were one or 2tb. #5 was 3tb. Didn’t hit a 4tb till #7 and the next one was a 16tb iron wolf pro.

[Marinatall_Ironside]

This technology sounds really cool. Still can't comprehend what it's all about, still curious about the sequential read/write speeds for it though.

[leadeater]

13 hours ago, Mira Yurizaki said:

Having two independent actuators may not decrease fault tolerance. It's not confirmed the two sets of heads will actually run in a RAID0 style, but may run in JBOD or something similar. In any case, having two independent actuators means if one is toast, the drive is still useful, just that half the platters are not. So depending on how the data is laid out when one group dies, some data my not be lost and you can still repurpose the drive for other uses until a replacement can be made.

I think people are mostly just confused over the wording used. These don't (I'm certain but could be wrong, this is not fact) show up as multiple disks that, statement was around the single disk performing as if it were two, hence appears to the application as being serviced by two disk performance wise. It's no different to saying an application sees and SSD as if it were 10 HDDs, performance wise. It's still one physical addressable device.

 

There's a very good reason why you wouldn't make it appear as two disks, disk replacement and failure behavior. What happens if one actuator fails? Now you have to cleanly fail the data odd the 'second disk' or is the disks set to make both fail in that situation? What about data resiliency? How do you ensure that data is adequately spread across disk slots to maintain correct failure domains.

 

Above issues are all solvable but it doesn't seem likely to me that you would have a single disk appear as two just because of the technical limitations and risks that would introduce. All anyone really cares about is the performance and two independent actuators that can read and write independently gives this performance increase and solves the most performance limiting scenario for HDDs, simultaneous  read and writes to the same disk. Simultaneous read/write is crippling, worse than halved performance as the seeking is compounding to the performance loss, this alone is enough reason to jump over to dual actuator disks, for anything not archival.

 

There's a very minor increase in failure chance just by the nature of having two motors to drive actuators over one before, both those very rarely fail.

 

Edit:

Never mind, for Microsoft these are two disk devices in one 3.5" disk bay.

Quote

“That has been key to developing the first iteration of the MACH.2 architecture, using dual LUNs with SAS — which fits Microsoft’s data center architecture and reasonably reflects what other large CSPs may do.”

For those that don't understand what a LUN is this is basically an identifier for a block device, every disk has these so having two means there are two block devices. This means they are seen as two literal different disks/block devices to the OS etc. A RAID array is a single LUN as an example of the more technical difference between a disk and a LUN, hence called block device not disk.

 

But there are different configurations coming.

Quote

As the Exos 2X14 dual-actuator hard drive program moves forward and expands, the drive family will be available in multiple flavors and SKUs to accommodate a full profile of enterprise use cases.

 

[justpoet]

Something of note is that the actuator movement is actually one of the most reliable things in the drive.  It is just magnetic positioning.

My only concern with this is really that the pivot point and bearings may cause more vibration than a single actuator head moving, depending on access patterns, even though it probably causes less in general use since half the mass is moving all at once.

[Mira Yurizaki]

4 hours ago, leadeater said:

There's a very minor increase in failure chance just by the nature of having two motors to drive actuators over one before, both those very rarely fail.

The increase in failure depends on how the system uses it. Look at airplanes, they duplicate a lot of critical features for the sole purpose of having a higher chance of survival.

[Doobeedoo]

Great way to improve speed really. More I/O per TB is important with increasing capacities. I also wonder how come higher RPM also didn't become more standard along. 

[Mira Yurizaki]

7 minutes ago, Doobeedoo said:

I also wonder how come higher RPM also didn't become more standard along. 

• Higher RPM is loud
• You can only spin something so fast before the centrifugal forces break the thing
• A lot of high RPM drives had less areal density than their slower spinning counterparts, which negated any bandwidth improvements. This is why the last WD VelicoRaptors used 2.5" platters.

Higher RPM was more for better seek times, which is what kills storage performance in a lot of use cases.

[Doobeedoo]

8 minutes ago, Mira Yurizaki said:

• Higher RPM is loud
• You can only spin something so fast before the centrifugal forces break the thing
• A lot of high RPM drives had less areal density than their slower spinning counterparts, which negated any bandwidth improvements. This is why the last WD VelicoRaptors used 2.5" platters.

Higher RPM was more for better seek times, which is what kills storage performance in a lot of use cases.

Yeah, I thought about it, the less areal density sure is one issue. More power and heat too, so probably not a good trade for it's benefits. I guess certain cases sure. 

I remember reading about 20 000 RPM prototype one time, though it never was released. 

[Mira Yurizaki]

11 minutes ago, Doobeedoo said:

Yeah, I thought about it, the less areal density sure is one issue. More power and heat too, so probably not a good trade for it's benefits. I guess certain cases sure. 

I remember reading about 20 000 RPM prototype one time, though it never was released. 

I also recall someone doing a test where they took a 7200 RPM drive, modified the firmware to make the heads only access the equivalent of a 2.5" platter, and found they got performance almost on par with a 3.5" 10K RPM drive

[SADS]

I really don't get why this, or more than 2 actuators, wasn't done sooner if HDD performance was such a problem.

This could have been done year ago?

[Rohith_Kumar_Sp]

Double the cost for head replacement. 

[Kilrah]

23 hours ago, redbread123 said:

i am so dumb, I never knew HDD had more than one "disc"

 

23 hours ago, NumLock21 said:

They do comes in multiple platters too, 4 is the most I've seen, but it depends on the size.

Apparently my 16TB Seagate Ironwolfs have 9 platters.

  

12 hours ago, Bombastinator said:

I don’t know how this tech would do much for consumer drives.  They’re mostly single platter. Especially for laptops.

Not true at all. 2.5" drives typically still use 1TB platters, so anything above that uses multiple ones. My 5TB 2.5" ones have, you guessed it, 5 platters. For 3.5" drives it seems to be around 1-1.5TB/platter max usually. I've taken apart many dead drives, and I've never come across a single-platter 3.5" model. Since there is enough height in 3.5" drives it makes little sense to put a single dense platter in a smaller capacity drive, easier and cheaper to put multiple less dense ones. Last I opened was a 2TB 3.5" one and it has 3 platters.

[Mira Yurizaki]

18 minutes ago, SADS said:

I really don't get why this, or more than 2 actuators, wasn't done sooner if HDD performance was such a problem.

This could have been done year ago?

I'm going to guess that a lot of hardware manufacturers simply incrementally improve what's current until their customers demand more and they've hit a wall.

[cj09beira]

On 12/5/2019 at 4:52 PM, porina said:

I struggled to see much advantage to this when it was first shown, and I still don't get it now. It's essentially two hard disks in one box, with the advantages and disadvantages that comes with it. There's still one head per side, so the only advantage is that the heads on each half of the drive can move independently of each other? I think having more than one head per side would help more, both in increasing throughput per platter and seek times.

this is an improvement, but i agree with you, they just have to move the disk down, and place another head on the top of the drive 

[NumLock21]

Wasn't this already a thing from a while back? I remember seeing it before.

[TechyBen]

13 hours ago, Bombastinator said:

I don’t know how this tech would do much for consumer drives.  They’re mostly single platter. Especially for laptops.  I vaguely recall some company or other in the past actually putting more than one drive arm opposite each other (so 2 read/write heads a side) on a single platter in an attempt to increase speed.  I don’t remember how it turned out.

Huh, I stripped a lot of 500gb and less drives out this summer. Old drives mind. Most had multi platters.

[Bombastinator]

21 minutes ago, TechyBen said:

Huh, I stripped a lot of 500gb and less drives out this summer. Old drives mind. Most had multi platters.

Yeah.  The older the drive the more platters it has for smaller size and the more expensive per gb it is.  Disk packs had 10 platters

https://en.m.wikipedia.org/wiki/Disk_pack