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

[williamcll]

Update: the drives are demonstrated at CES on their Lyve Drive platform: https://www.anandtech.com/show/15358/seagate-demonstrates-hamr-and-dualactuator-hard-drives-in-the-lyve-drive-mobile-system

 

Multi-actuators have been one of the methods for Hard disks to catch up to SSD's performance, Seagate demonstrated it a year back and now it is finally on production for enterprise use.

Quote

Seagate has a new HDD that leverages its MACH.2 multi-actuator technology, the Exos 2X14. While the technology was announced two years ago with the promise of doubling throughput, Seagate, along with its customer Microsoft, are able to show results of the technology and its performance breakthroughs. The technology is beneficial to applications such as content delivery networks (CDNs), video streaming, mail servers, backup/shuttle services, Hadoop, and cloud applications and is being tested for data center deployment now. 

As we said at the initial announcement, Seagate’s solution to this is simple enough; if the number of actuators limit performance, add more. For its first generation of the new product, Seagate will be adding one actuator for two total on a single pivot point. According to the company, half the drive’s recording heads will operate together as a unit, while the other half will operate independently as a separate unit. This enables a hard drive to double its performance while maintaining the same capacity as that of a single actuator drive.

 

The MACH.2 technology should also prevent something referred to as stranded capacity. In a nutshell, the capacity of HDDs is increasing faster than the drives ability to transfer the data. The workaround has been to put less data on the drives to hit the maximum throughputs, thus “stranding” capacity. MACH.2 can solve this issue. An example that Seagate stated was a data center can install a 14TB dual-actuator drive in each slot. To the application, it looks very similar to the high throughput provided by two 7TB drives — but it fits into a single slot.

Microsoft has been a close collaborative partner during the development of MACH.2 multi-actuator technology. Microsoft’s main goal is to leverage the technology to maintain the IOPS required for some of Microsoft’s cloud services including Azure and the Microsoft Exchange Online email service, while increasing available storage capacity per data-center slot. Microsoft stated that it has completed its initial round of testing that included a functional deployment testing of the Exos 2X14, including full infrastructure testing for interoperability and compatibility with the data center infrastructure, product robustness, reliability and ease of integration with its Project Olympus system architecture — as well as the tests on the performance increase MACH.2 provides. Microsoft used its Microsoft Exchange Server Jetstress tool, which simulates Exchange 2013 and Exchange 2016 disk I/O load on a server to verify the performance and stability of a disk subsystem. Microsoft saw that they were getting close to twice the throughput and IOPS with the new technology.

Seagate also believes that MACH.2 can solve issues like IOPS/TB and cost/TB. The technology will allow for twice the throughput and IOPS at higher capacities than other HDDs while taking up the same footprint. If the use cases line up right and the price is right, this can be used as a potential SSD alternative for cloud customers. 

Source:https://www.storagereview.com/seagate_exos_2x14_doubles_iops_for_microsoft/
https://www.cxotoday.com/press-release/microsoft-nearly-doubles-iops-using-seagate-exos-with-mach-2-dual-actuator-technology/
https://mp.weixin.qq.com/s/tO0zt_bWs99-IOtvFcVMHA
Thoughts: It will be many more months before this technology will be available in the consumer market, and if it isn't as cheap as QLC/PLC SSDs I doubt it would be competitive. Would be nice to hear from the seagate forum affiliate about this.

[The King of the Undead]

interesting. still has the drop death problem but it may be cheaper and if you drop your desktop more expensive things are going to die anyway.

[SansVarnic]

I find this very interesting, I am still not 100% comfortable about storing long term on ssds at the moment but his may be something to look into. Additionally probably the first innovation in HDD tech to come in a while, not that it is a leap or anything but a different angle being taken.

[porina]

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.

[matrix07012]

Pretty cool. I wonder what's the failure rate going to be like, since they added a new point of failure.

[RejZoR]

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.

[redbread123]

1 hour ago, williamcll said:

Multi-actuators have been one of the methods for Hard disks to catch up to SSD's performance, Seagate demonstrated it a year back and now it is finally on production for enterprise use.

Source:https://www.storagereview.com/seagate_exos_2x14_doubles_iops_for_microsoft/
https://www.cxotoday.com/press-release/microsoft-nearly-doubles-iops-using-seagate-exos-with-mach-2-dual-actuator-technology/
https://mp.weixin.qq.com/s/tO0zt_bWs99-IOtvFcVMHA
Thoughts: It will be many more months before this technology will be available in the consumer market, and if it isn't as cheap as QLC/PLC SSDs I doubt it would be competitive. Would be nice to hear from the seagate forum affiliate about this.

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

[NumLock21]

25 minutes ago, redbread123 said:

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

They do comes in multiple platters too, 4 is the most I've seen, but it depends on the size. Taking a 1TB drive as an example, they can have a single platter at 1TB or 2 platters at 500GB each. The single platter version is usually faster, because of the areal density of the platters.

[Velcade]

2 hours ago, williamcll said:

Thoughts: It will be many more months before this technology will be available in the consumer market, and if it isn't as cheap as QLC/PLC SSDs I doubt it would be competitive. Would be nice to hear from the seagate forum affiliate about this.

It's not meant to compete with SSDs

 

Quote

An example that Seagate stated was a data center can install a 14TB dual-actuator drive in each slot. To the application, it looks very similar to the high throughput provided by two 7TB drives — but it fits into a single slot.

Data Centers and home storage are where you'll see this take off. 

[RejZoR]

3 minutes ago, NumLock21 said:

They do comes in multiple platters too, 4 is the most I've seen, but it depends on the size. Taking a 1TB drive as an example, they can have a single platter at 1TB or 2 platters at 500GB each. The single platter version is usually faster, because of the areal density of the platters.

Startup times are also longer on multiplatter drives. I've had a WD Caviar Black 2TB that had 4 platters (each 500GB, I think it was the first 2TB drive released back then). It took quite few seconds for the drive to spin up that huge mass of platters up to operational speeds.

[TigerHawk]

1 hour 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.

Data striping. You can have a raid 0 setup within the same hard drive. The top actuator reads/writes half the data while the bottom one reads/writes the other half. That's where the doubling comes in. You are literally utilizing the data twice as fast. This is exactly how raid 0 functions. You could do raid 10 with only two hard drives. 

 

The firmware has to support such a thing though but from what I gather that seems to be the Idea from the start. 

[Mira Yurizaki]

2 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.

It can service two different data requests at the same time, assuming the data lives on separate halves. So at best you're looking at double the performance, at worse, as much as a single actuator. And it's likely the firmware will make sure data doesn't mostly live on one half or the other.

[porina]

6 minutes ago, TigerHawk said:

Data striping. You can have a raid 0 setup within the same hard drive. The top actuator reads/writes half the data while the bottom one reads/writes the other half. That's where the doubling comes in. You are literally utilizing the data twice as fast. This is exactly how raid 0 functions. You could do raid 10 with only two hard drives. 

You wouldn't need the actuators to move independently to "raid 0" it though. You can simply read from multiple heads at the same time. Why not read/write to all heads at the same time?

 

2 minutes ago, Mira Yurizaki said:

It can service two different data requests at the same time, assuming the data lives on separate halves. So at best you're looking at double the performance, at worse, as much as a single actuator. And it's likely the firmware will make sure data doesn't mostly live on one half or the other.

This is more in line with my thinking, but this would only offer a marginal latency improvement.

[NumLock21]

4 minutes ago, RejZoR said:

Startup times are also longer on multiplatter drives. I've had a WD Caviar Black 2TB that had 4 platters (each 500GB, I think it was the first 2TB drive released back then). It took quite few seconds for the drive to spin up that huge mass of platters up to operational speeds.

WD Black runs at 7200RPM so a few seconds to spin up to operational speeds isn't a huge deal.

[MichaelWd]

Based on the article, the advantage of this is more capacity per slot. So saving space. And I guess also saving data/power connections.

 

So for a given number of servers, this gives you a higher capacity ceiling while maintaining acceptable speed.

[Mira Yurizaki]

27 minutes ago, porina said:

This is more in line with my thinking, but this would only offer a marginal latency improvement.

Skimming the marketing, it doesn't look like it's for latency. It's for throughput.

[SpookyCitrus]

More moving parts more failure points, I'm interested to test one of these out. Maybe have one half of the drive for files and the other as a back up? 

[MichaelWd]

Just now, SpookyCitrus said:

More moving parts more failure points, I'm interested to test one of these out. Maybe have one half of the drive for files and the other as a back up? 

But not double the number of moving parts (platter spins as a whole, so one motor for that). So it is a per GB reduction in moving parts.

[SpookyCitrus]

4 minutes ago, MichaelWd said:

But not double the number of moving parts (platter spins as a whole, so one motor for that). So it is a per GB reduction in moving parts.

I never said double. I just said more... more read arms and more platters, aka more failure points. In my experience it's usually the read arm that mucks up the drive, whether it breaks, scratches a platter, or just stops working. Out of all the drives I've worked with that were failing and or failed only one of them had the platter motor go out.

[RejZoR]

1 hour ago, NumLock21 said:

WD Black runs at 7200RPM so a few seconds to spin up to operational speeds isn't a huge deal.

Bootups, waking from sleep and recovery from crashes was quite noticeable. I ran it 24/7 anyway so it didn't bother me all that often, but the spinup time was quite noticeable compared to older drives that had less platters. I think I had 750GB Samsung F1 before this WD Caviar Black that had 2x 375GB platters iirc and it was way faster at spinning up. But yeah, all that is past for me. Running SSD's exclusively in all my devices for years now and boy it's like living in a future No lag, no scratching and grinding, just pure silence and speed.

[Mira Yurizaki]

33 minutes ago, SpookyCitrus said:

I never said double. I just said more... more read arms and more platters, aka more failure points. In my experience it's usually the read arm that mucks up the drive, whether it breaks, scratches a platter, or just stops working. Out of all the drives I've worked with that were failing and or failed only one of them had the platter motor go out.

I'm pretty sure the only thing they did was split the arms into two groups and have an actuator on each so the groups can move around independently.

 

The image is very likely not representative of any actual drive. Besides that, hard drives come in a variety of different platter/head numbers.

[jagdtigger]

4 hours ago, porina said:

I think having more than one head per side would help more, both in increasing throughput per platter and seek times.

Yeah but that wouldnt fit into the standard size.... This  way they can keep the traditional size but loose one platter in capacity (AFAIK they use one side of a platter to store data needed for the controller to actually position the head above track and stuff).

Edited December 5, 2019 by jagdtigger

[mr moose]

On the scale that data centres work this drive would make huge sense,  almost twice the storage density plus all the perks in extra speed.

[GDRRiley]

now all we need is backblaze to have a bunch for a year to tell us how much they fail

[Unixsystem]

2 hours ago, MichaelWd said:

Based on the article, the advantage of this is more capacity per slot. So saving space. And I guess also saving data/power connections.

 

So for a given number of servers, this gives you a higher capacity ceiling while maintaining acceptable speed.

A semantic point, but it's less "more capacity" and more that it's "more capacity at a reasonable performance".

 

A full 10+TB drive is going to take ages to seek to completely different parts of the disk, so some data centera may deliberately only use a small fraction of the disk in order to maintain usage performance. With multiple heads, they can use all/a larger percentage of the disk without meaninful performance loss.