Seagate's HAMR to drop in 2020: Multi-actuator disk drives on the way In 2020 Seagate will introduce its first multi-actuator disk drives using Heat-Assisted Magnetic Recording (HAMR) tech with 20TB capacities. This was revealed at the Open Compute Project summit, which took place yesterday and today at the San Jose Convention Center. Wells Fargo analyst Aaron Rakers was there and distributed …
I like your previous article's quotes from Seagate:
"Half the drive's recording heads will operate together as a unit, while the other half will operate independently as a separate unit,",
and
"The host computer can treat a single Dual Actuator drive as if it were two separate drives,"
And so what you've got is one disk enclosure, and one interface, but effectively two independently operated drives. Not two independent read/write arms accessing the same entire disk. To access a particular set of data, you are still waiting on the queue of commands for a single arm. Ho hum.
We've got to look for something like the additional word 'parallel' to get the nirvana of "times N" access to one set of data. Though duplication of data using RAID may help? Oh, that's reads only.
Even on a single actuator disk, only one head is reading or writing at the same time, no matter the number of platters. With multiple actuators, two data requests can be made at the same time, full parallelism. This is a pretty big breakthrough and definitely paves the way for more tech like it.
And what exactly is wrong with having what is effectively two independent drives? You get X IOPS per drive, so having two drives (whether two physical drives or two "effectively independent" drives) gives you 2X IOPS. Just because you don't have two heads able to access the same block of data isn't a bad thing. Servers always have a lot of outstanding requests, so they'll be able to get through the queue 2X as quickly as they could otherwise.
Remember, this isn't targeted at home users who will generally have little or no parallelism in their I/O requests. Those people are going to be using SSDs, not 20 TB HAMR drives.
With 32 TB SSDs already shipping, HAMR looks to be a sideshow. Heck, we'll get 100 TB 3.5 inch SSDs by Xmas. The best we can hope for with HAMR is very slow IO and 10X the number of racks to hold a given capacity.
Yes, SSD will cost more per terabyte, but that 10X factor in appliance count offsets much of that!
True, but if it can "race to idle" the average power draw may be close to the idle power rather than active. Even the time to resume from full device sleep is lower than the average access time of a spinning disk.
The question becomes how fast it can recognize that it will be idle, and what the proportion is. Hard disks may remain a good choice for steady but not huge workloads over large contiguous data. Which includes many file serving operations.
"Yes, SSD will cost more per terabyte, but that 10X factor in appliance count offsets much of that!"
I agree and there's an additional aspect to consider with these new heat assisted hard drives. In effect, making these drives more complex also means that there's a greater potential for things to go wrong and I'd want to see independent reliability statistics before even considering deploying these new hard drives.
Well, you can't blame the HDD manufacturers for trying to stay in the game. HAMR HDDs have been in development for more than a decade, and you still can't buy one. MAMR HDDs are exotic, and you won't see them for sale for a long time either. HDDs with dual actuators have been tried before and abandoned. Maybe it can be made to work, but adding electro-mechanical complexity to HDDs can only be a source of new HDD failures. In the meantime, caging electrons in NAND flash is a quicker way to increase capacity dramatically. Niche markets for HDDs will hang on for maybe another ten years, but when those use cases can be challenged by NAND flash storage, it will be the end for HDDs. The storage imperative in the 21st century is maximum capacity with minimum power requirements. Anything that does not do that will not be around.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
