Will multi-actuators save the disk drive?
In a blog post - republished in Storage Newsletter - by Seagate's Jason Feist, he says that new multi-actuator technology will allow hard disk drives (HDDs) to meet hyperscale data center requirements, by
. . . solving this concurrent need for increased performance by enabling parallelism of data flows in and out of a single hard drive through Multi Actuator technology, which offsets the potential usability issues of super-high-capacity hard drives by doubling hard drive performance.
Two heads are better than one
The multi-actuator technology (MAT) doesn't increase the number of heads reading and writing. What it does is essentially package two disk drives in to one standard 3.5" form factor. One set of actuators reads and writes to half the platters, and the other set handles the remaining platters.
In theory, then, the two logical drives could be set to mirror each other, or act as two HDDs in a RAID 0 array. In the latter case, the sequential performance of the drive should indeed double.
But a RAID 0 can also be configured to increase IOPS, by setting the chunk size so that most I/Os are serviced by a single actuator. That should, depending on workload, double the IOPS.
Why now?
The multiple actuator idea is not new. In fact, drum drives, a predecessor to disk drives, had a read/write head for every track for massive parallelism. But R/W heads are expensive, so the few companies that tried to market multi-head drives over the years have failed.
In the post, Feist also points to the 2019 arrival of HAMR (heat-assisted magnetic recording) drives, which promise much higher data densities and capacities, as a reason for MAT. There will be so much capacity that single actuators won't be fast enough to move all the data the drives are capable of storing.
He also refers to this as Seagate's first generation of MAT. Potentially, later models could offer three or more independent actuators, increasing performance as RAID 0 does, or improving data integrity as RAID 5 does.
The Storage Bits take
Disk drive volumes have been hit by the move of notebooks and some desktops (iMac Pro, for example) to SSD-only configurations. Those consumer markets are never coming back.
Servers will continue to use disk drives because they are much cheaper than SSDs for bulk data storage. With the advent of even higher density storage, MAT will become a common feature.
But will high density HDDs actually arrive? Feist's comment that HAMR drives will arrive in volume in late 2019 means that the timeframe remains uncertain. Two years is a lifetime in high-tech product development. We'll have to wait and see if Seagate actually samples HAMR drives in 2018 before we can have any confidence that super-high capacity drives are actually coming in 2020.
Courteous comments welcome, of course.