Thursday, March 17, 2016

Dr. Pangloss loves technology roadmaps

Its nearly three years since we last saw the renowned Dr. Pangloss chuckling with glee at the storage industry's roadmaps. But last week he was browsing Slashdot and found something much to his taste. Below the fold, an explanation of what the good Doctor enjoyed so much.

The Slashdot entry that caught the Doctor's eye was this:
Several key technologies are coming to market in the next three years that will ensure data storage will not only keep up with but exceed demand. Heat-assisted magnetic recording and bit-patterned media promise to increase hard drive capacity initially by 40% and later by 10-fold, or as Seagate's marketing proclaims: 20TB hard drives by 2020. At the same time, resistive RAM technologies, such as Intel/Micron's 3D XPoint, promise storage-class memory that's 1,000 times faster and more resilient than today's NAND flash, but it will be expensive — at first. Meanwhile, NAND flash makers have created roadmaps for 3D NAND technology that will grow to more than 100 layers in the next two to three generations, increasing performance and capacity while ultimately lowering costs to that of hard drives."Very soon flash will be cheaper than rotating media," said Siva Sivaram, executive vice president of memory at SanDisk.
ASTC 2015
The article by Lucas Merian that sparked the post has the wonderfully Panglossian title These technologies will blow the lid off data storage, and it has some quotes that the Doctor really loves, such as the one above from Siva Siaram, and the ASTC technology roadmap showing a 30% Kryder rate from next year with HAMR and BPM shipping in 2021.

But the good Doctor tends not to notice that the article is actually careful to balance things like "HAMR technology will eventually allow Seagate to achieve a linear bit density of around 10 trillion (10Tbits) per square inch" with "Seagate has already demonstrated HAMR HDDs with 1.4Tbits per square inch" (my emphasis). If you pay attention to these caveats it is actually a good survey of the technologies still out on the roadmap.

But curmudgeons like me remember that back in 2013 the Doctor was rubbing his hands over statements like:
Seagate is projecting HAMR drives in 2014 and WD in 2016.
In 2016 we hear that:
Seagate plans to begin shipping HAMR HDDs next year.
Seagate 2008
So in three years HAMR has gone from next year to "next year". Not to mention the graph I keep pointing to from 2008 showing HAMR taking over in 2009 and BPM taking over in 2013. So actually HAMR has taken 8 years to go from next year to next year. And BPM has taken 8 years to go from 5 years out to 5 years out.

Why is this? As the technologies get closer and closer to the physical limits, the difficulty and cost of moving from "demonstration" to "shipping" increase. For example, lets suppose Seagate could demonstrate HAMR in 2013 and will ship it in 2017. BPM is even harder than HAMR, so if it is going to ship in 2021 it should be demonstrable this year. Has anyone heard that it will be?

The article also discusses 3D NAND flash, which also featured in Robert Fontana's wonderful presentation to the Library of Congress Storage Architecture workshop. From his slides I extracted the cost ratio between flash and hard disk for the period 2008-2014, showing that it was converging very slowly. Eric Brewer made the same point in his FAST 2016 keynote. Flash is a better medium than hard disk, so even if the manufacturing cost per byte were the same, the selling price for flash would be higher. But, as the article points out:
factories to build 3D NAND are vastly more expensive than plants that produce planar NAND or HDDs -- a single plant can cost $10 billion
so no-one is going to make the huge investment needed for 3D NAND to displace hard disks from the cloud storage market because it wouldn't generate a return.

The article also covers "Storage Class Memories" (SCM) such as Intel/Micron's Xpoint, mentioning price:
even if Intel's Xpoint ReRAM technology enters the consumer PC marketplace this year, its use will be limited to the highest-end products due to cost.
Intel 2015
Actually, Intel isn't positioning Xpoint as a consumer storage technology but, as shown in the graph, as initially being deployed as an ultra-fast but non-volatile layer between DRAM and flash,

As I commented on Fontana's presentation:
The roadmaps for the post-flash solid state technologies such as 3D Xpoint are necessarily speculative, since they are still some way from shipping in volume. But by analogy with flash we can see the broad outlines. They are a better technology than flash, 1000 times faster than NAND, 1000 times the endurance, and 100 times denser. So even if the manufacturing cost were the same, they would command a price premium. The manufacturing cost will initially be much higher because of low volumes, and will take time to ramp down.
So, despite the good Doctor's enthusiasm, revolutionary change in the storage landscape is unlikely. We are unlikely to see ASTC's 30% Kryder rate, 3D NAND will not become cheaper for bulk storage than hard disk, and SCM will not have a significant effect on the cost of storage in the foreseeable future.

1 comment:

David. said...

Brian Hayes Where’s My Petabyte Disk Drive? is worth reading, if a bit naive about the forces driving storage technology.