Subject: Storage | December 9, 2017 - 11:46 PM | Tim Verry
Tagged: PMR, toshiba, helium, Hard Drive, enterprise, cmr, cloud storage, 14tb
Toshiba recently took the wraps off of a new hard drive series aimed at the enterprise market. What makes the MG07ACA series interesting is that Toshiba is offering a 14 TB 3.5” drive without resorting to using Shingled Magnetic Recording. Instead, the new MG07ACA series uses standard recording methods (CMR) and nine ~1.556 TB PMR (perpendicular magnetic recording) platters in an helium filled hermetically sealed enclosure to hit 40% more capacity and up to 50% better power efficiency than the previous MG06ACA (10 TB) series. The new drives are also important because they represent the first foray into helium filled hard drives for Toshiba following the company pushing air breathing drives to the limit with its seven platter models.
The new drives are standard 7200 RPM models with 256 MB of cache and a SATA 6 Gbps interface. The 14 TB model is able to hit 260 MB/s sustained transfer while the slightly lower areal density of the 12 TB model puts it at a 250 MB/s transfer speed maximum. They are able to hit 167 random 4K read IOPS and 70 random 4k write IOPS (which is fun to compare to even the slowest SSDs today, but these drives aren't for random workloads). Toshiba rates the drives at a fairly industry standard 550 TB per year workload and 2.5 million hours MTBF with a five year warranty. Toshiba is reportedly using its own laser welding technology to seal the drives and keep the helium contained. The MG07ACA drives are offered in emulated 512 (512e) and 4k native sectors with the 512e models featuring Toshiba Persistent Write Cache technology to prevent data loss in the event of power failure while the drives are executing read-modify-write operations. The power loss protection (PLP) is important for enterprise customers using these drives to upgrade the storage in their legacy software and hardware setups.
The MG07ACA series includes 14 TB 9-disk and 12 TB 8-disk drives. That’s a lot of platters in a single drive, but Toshiba claims that going this route with CMR / PMR reduces the total cost of ownership (TCO) for enterprise customers that are buying up high capacity drives for their cloud storage and big data storage needs. The drives are allegedly more power efficient and trusted in the enterprise market as opposed to the newer shingled drives. I suppose these drives are also useful as they can be drop in upgrades of lower capacity models.
John Rydning, Research Vice President for hard disk drives at IDC was quoted in the press release in saying:
"While enterprise server and storage customers realize that shingled magnetic recording (SMR) technology can improve HDD capacity, the adoption of SMR HDD products into server and storage systems is a transition that will take several years,"
Interestingly the drives offer 1.5 TB / platter in the 12 TB model and a bit more than 1.55 TB / platter in the 14 TB drive. With SMR technology hitting up to 1.75 TB / platter so far, using that could get a 14 TB drive with just 8 platters, but that is still fairly close that I suppose going with the longer track record of non shingled PMR and its reliability is more important to the enterprise customers.
In order to cram 9 platters into a standard 3.5" drive, Toshiba had to make the platters thinner and move to helium instead of air. Specifically, Toshiba is using 0.635mm Showa Denko (SDK) PMR platters that are a mere 1.58mm apart! The drives have Nidec motors on the top and bottom as well as environmental sensors and RVFF (Rotation Vibration Feed Forward) vibration compensation technology which is important when you have nine platters spinning at 7200 RPM in each drive and then hundreds of drives are placed in close proximity to each other in server racks and SANs. The move to helium and thinner platters is a big part of the power savings in this drive with the platters being easier to spin up and exhibiting less flutter moving through the much less dense helium versus air. Toshiba claims that the MG07ACA series uses up to 7.6 watts in normal operation and 4.6 watts at idle (0.32W/GB).
According to AnandTech, Toshiba will begin sampling the new hard drives later this month and will sell the drives to its large enterprise customers within the first half of next year. Once demand from the big data crowd has been met, Toshiba will being selling the drives through distributors which means enthusiasts will be able to get their hands on the drives through normal channels by the end of 2018. Exact pricing and availability have not been announced at this time.
- Western Digital Launches 14TB Enterprise Hard Drive for Big Data
- Western Digital Launches 12TB Gold Hard Drive To Consumers
- WD and HGST Refresh Enterprise SSDs to Include 8TB, Push HDDs to 12TB and Beyond
- Western Digital MAMR Tech Pushes Future HDDs Beyond 40TB
- Seagate BarraCuda Pro 10TB Review - Massive Helium Client HDD
Subject: Storage | October 11, 2017 - 11:16 PM | Allyn Malventano
Tagged: western digital, wdc, WD, STO, Spin Torque Oscillator, SMR, PMR, Microwave Assisted Magnetic Recording, microwave, MAMR, HAMR, FMR
Today Western Digital made a rather significant announcement in the field of HDD technology. We’ve previously talked about upcoming ways to increase the density of HDD storage, with the seeming vaporware Heat Assisted Magnetic Recording (HAMR) forever looming on the horizon, just out of reach.
WD, like others, have been researching HAMR as a possible way of increasing platter densities moving forward. They were even showing off prototypes of the technology back in 2013, but a prototype is a far cry from a production ready, fully reliable product. Seagate had been making stronger promises of HAMR, but since we are already 5 years into their 10-year prediction of 60TB HAMR HDDs (followed by further delays), it's not looking like we will see a production ready HAMR HDD model any time soon.
Ok, so HAMR is not viable for now, but what can we do? Seems WD has figured it out, and it's a technology they have been kicking around their labs for nearly a decade. Above we see the PMR limit of ~1.1 Terabits/square inch. SMR pushes that figure to 1.4, but we are running up against the so-called 'writeability limit', which is the point at which the write head / magnetic field is too small to overcome the paramagnetic threshold of the smaller magnetic domains of higher density media. We are used to hearing that the only way to raise that limit was to heat the media with a laser while writing (HAMR), but there is a different / better way - Microwave Assisted Magnetic Recording, or MAMR for short.
Don't let the 'microwave' part of the term fool you - we are not microwaving the media with sufficient energy to actually heat it. Instead, we are doing something *way* cooler. The slide above shows how smaller grain size (higher density) requires a stronger write field to reach sufficient energy levels to reliably store a bit of data. Now check out the next slide:
This is a lot to grasp but allow me to paraphrase greatly. Imagine a magnet with a north and south pole. If you came along with a stronger magnet and attempted to reverse its polarity by directly opposing the currently stored state, it's generally difficult to do so. Current HDD tech relies on the field being strong enough to overcome the stored polarity, but MAMR employs a Spin Torque Oscillator, which operates at a high enough frequency (20-40 GHz) to match the ferromagnetic resonance of the media. This causes a precession of the stored field (like a gyroscope) and tilts it about its vertical axis. This resonance adds the extra energy (in addition to the write field) needed to flip the field to the desired direction. What's amazing about this whole process is that thanks to the resonance effects, the STO can increase the effectiveness of the write field 3-4x while only consuming ~1/100th of the power compared to that needed to generate the write field. This reduction in the damping constant of the media is what will enable smaller magnetic domains, therefore higher platter densities in future MAMR-equipped HDDs.
One of the best things about this new tech is that it is just a simple addition to all of other technologies already in place today. Western Digital was already making their drive heads with an advanced 'damascene' process, silently introduced about three years ago. To oversimplify the description, damascene is a process that enables greater physical precision in the shape of the head, which helps increase density. What makes this process a bigger deal now is that it more easily enables integration of the Spin Torque Oscillator into the head assembly. Aside from this head-level change and another pair of leads to provide a very small drive current (~1-2mA), every other aspect of the drive is identical to what we have today. When it comes to a relatively radical change to how the writing can be accomplished at these upcoming higher densities, doing so without needing to change any of the other fundamental technologies of the drive is a good thing. By no change, I really mean no change - MAMR can be employed on current helium-filled drives. Even SMR.
Western Digital also slipped in another announcement, which is the shift from the older style 'nested actuator' (introduced with 2TB HDDs back in 2009), to a newer 'micro-actuator'. The newer actuator moves the articulation point much closer to the head compared to the previous technology, enabling even finer head tracking, ultimately resulting in increased track pitch. WD currently sits somewhere around 400 tracks per inch (TPI), but they hope to reach 1 million (!) thanks to this new tracking combined with MAMR and improved media chemistry.
Now this doesn't mean we will see a sudden influx of 40TB HDDs hitting the market next week. WD still has to scale up production of STO-enabled heads, and even after that is complete, the media technology still needs to catch up to the maximum capabilities of what MAMR can achieve (creating smaller magnetic domains on the disk surface, etc). Still, it's nice to know that there is a far simpler way to flip those stored bits around without having to resort to HAMR, which seems to be perpetually years away from production. Speaking of which, I'll leave you with WD's reliability comparison between their own HAMR and MAMR technologies. Which would you choose?
Oh yeah, and about that supposed SSD vs. HDD cost/GB crossover point. It may not be as soon as we previously thought:
Full press blast appears after the break.