Thankfully, the Shingled Magnetic Recording technology I discussed in my first post on SMR can provide the next capacity boost the drive industry needs. Unfortunately, the challenges of SMR technology, which I wrote about in my article on using SMR drives will both limit just how much added capacity SMR can deliver and prevent SMR from being used for any sort of random workload. However, a technology that generates plenty of hardware store and worksite jokes -- Heat Assisted Magnetic Recording (HAMR) -- promises to help.
As I wrote previously, disk drive density -- and therefore capacity -- has stalled because the current media has reached the point where if the magnetic grains are made any smaller, they’ll no longer hold their state. Using a magnetic medium with higher coercively would allow smaller magnetic domains, but would require magnetic field strengths higher than the heads can “focus,” to bits at over 1TB/square inch.
HAMR drives use a magnetic medium that requires a field of 50,000 oersted to flip a domain from north to south or vice versa. That’s 10 times the 5,000 oersted field today’s PMR (Perpendicular Magnetic Recording) drives use. Faced with material with such a high coercivity and heads that couldn’t generate a strong enough field to write to it, drive engineers went the Dr. Evil route and put a laser on the head to heat the media before it writes to it.
The laser is a few milliwatts, about as powerful as the one in a good laser pointer, which is a small fraction of the power used by the spindle motor, meaning HAMR drives won’t actually run hotter than their PMR predecessors.
The laser heats the medium to a temperature above its Curie point, where all the magnetic domains randomize, about 800 degrees Fahrenheit. As the material cools -- a process that by the way takes under a nanosecond -- its coercivity increases as the temperature decreases. At some intermediate temperature, it’s at a point where it requires a field about twice as strong as a PMR drive and rapidly (did I say less than one nanosecond?) increases in magnetic stability.
[Read why Howard Marks doesn't think it's likely that optical storage will make a big comeback in "Can Facebook Bring Back Optical Storage?"]
Actually, building HAMR drives requires advancements in several technologies. The engineers at Western Digital and Seagate have to figure out how to mount the little laser on the head while maintaining a constant flying height and a constant aim point a few micrometers in front of where the data is being written. Then, they have to engineer the substrate to conduct heat away from the lasered spot very quickly at a predictable rate. They are addressing this problem with glass substrates and a sputtered layer that conducts the heat horizontally, so it’s at the right temperature when it rotates under the heads.
Since the whole heat, write, cool process takes so little time, HAMR isn’t limited to sequential workloads like SMR drives, but will boost capacity for 15K RPM drives running random workloads as well.
The folks at the Advanced Storage Technology Consortium, which is made up of the drive manufacturers and their suppliers like LSI, Marvell and Xyratex, are working together to bring both HAMR and bit patterned media to market. As long as some gremlin doesn’t raise its ugly head to throw development off the roadmap, HAMR should let the vendors boost density from today’s 800GB/in2 to about 5TB/in2 which all other things being equal, would mean a 25TB disk drive.
While non-volatile memory based solutions will continue to take over the performance end of the storage market, HAMR and shingled drives will ensure that the spinning disk will be the most economical place to store data online.