Hitachi Data Systems CTO Hu Yoshida sees no reason for anyone to buy hard disk drives for performance reasons now...
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
that flash is cost competitive.
Yoshida asserted that flash storage drives will replace 15K and 10K rpm hard disk drives (HDDs) this year, and he predicted that all purchasing will be done on flash, rather than disk within five years. In part one of the wide-ranging interview, Yoshida discussed how Hitachi storage needs to transform its business to compete in the changing and disrupted market.
What will be the most important storage trends in 2016?
Hu Yoshida: Flash is going to replace high-performance disks this coming year, because the price points have come down considerably. Our media device does compression. It's always on, because it's done in the media -- not in the control unit. If you consider just one-to-one compression, we are about 28% lower than a 15K hard drive over a five-year cost, including environmental and maintenance. This is on a bit capacity basis. ... If you consider two-to-one compression, flash is even cheaper than the 10K hard disk drives. And this is just our use of [multilevel cell flash]. We haven't gone to 3D [NAND flash], which will give us even more capacity and advantages and more reliability, because they're relaxing the cell size in 3D in terms of durability.
I don't see any reason for anybody to buy hard disks for performance in the coming year. ... There's even the thought of flash being used for a long-term archive. Having a flash drive would help to speed the [data] retrieval in a large archive. And concerns about durability are going away, especially if you have an archive that's infrequently accessed.
Do you think hard disk drives will die out?
Yoshida: I know people said, 'Tapes are going to die,' and they never did die out. But tapes always had the advantage of being an off-the-shelf storage medium. And the technology roadmap for tape is still very strong. The technology for the hard disk just sort of hit a wall.
The other thing is that what drives the cost down is volume. The volumes used to come from the commodity market. Now, most of the PCs and servers all have flash. So, there's less volume to drive the cost down. From an economic standpoint, it's not going to be viable to continue building hard drives, because those [HDD] plants are very expensive.
Will tape outlast HDDs?
Yoshida: I think you'll see disk fade before you see tape fade away. Of course, there is a huge inventory of disks out there, so that's not going to disappear overnight. And there's not enough capacity to build enough flash to store all the data we want to store. So, disks will be around for that reason. But as they age out, they're going to need to be replaced more with flash.
Hu YoshidaCTO, Hitachi Data Systems
How long will it take for disks to fade away?
Yoshida: I think within five years, probably all new purchasing will be done on flash, rather than disks.
Will there be only flash and tape?
Yoshida: No, some sort of nonvolatile memory. The flash drive is there for now, and flash drives enjoy the volumes from the commodity market, but I think there are other nonvolatile memories that are on the horizon. You saw Intel working with Micron to come out with this new type of technology that is really almost as fast as [dynamic RAM] and has more durability than flash. You don't have the write format requirement that you have in these new technologies.
Do you still see an advantage for Hitachi storage making its own flash modules?
Yoshida: Yes, a tremendous advantage, because in flash storage, most of the management of that flash cell has to be done in the media -- not up in the control unit, where you have no visibility into where cells are. And you need to have enough processing power in the media. We use a quad-core ARM processor there. We use multipathing. Things like housekeeping are taken out of the I/O path, so we don't suffer performance degradation when you go through the housekeeping. We are doing compression down in the media itself. If you do compression in the storage controller, you're going to use the software in there, and so you're going to have overhead that's going to impact all the devices attached to that controller.
There's also the possibility with that type of processing power to offload things like RAID into the drive instead of having to communicate back to the controller.
The [solid-state drive] SSD vendors could do the same thing. But they're building SSDs for the general market to be used in servers that don't do RAID, for instance. We build it specifically for the storage market. We know in storage, you often format the zeros. We don't have to write those zeros. We just have a pointer. So, there [are] a lot of things that we could do by having intelligence down in drives that you can't do in an all-flash array that's just been trying to do everything in a controller that has no visibility into the location of those cells and spares.
Hitachi storage launches new all-flash array
Hitachi storage strategy leans toward software
Hitachi storage focuses on Internet of Things