Original URL: http://www.theregister.co.uk/2010/07/16/coughlins_thoughts/

Of hybrid hard drives and capacity boosts

Coughlin's cogitations

By Chris Mellor

Posted in Storage, 16th July 2010 11:18 GMT

The extra platters and shingled writing future revealed by Xyratex CEO Steve Barber seems to have been influenced by Tom Coughlin of Coughlin Associates and his "Hard Disk Drive Capital Equipment Market and Technology Report".

Coughlin's reports are used by many people in the hard disk drive (HDD) production and buying communities. I thought talking direct to the man himself might be a good idea and so we had an email Q&A. Here it is, with Tom's replies in bold type:

El Reg: How do think hybrid hard drives will develop (flash + spinning disk in one enclosure)? Do they need host software changes? Which markets are they best suited for?

I think there will be more than one path pursued which will combine HDDs and flash memory.

One approach is the one Seagate [Momentus XT] is taking where they put the flash in the PCB [printed circuit board] of the HDD and use it for a non-volatile storage cache. This version differs from what they and other companies offered in 2007 in that it does not depend upon the operating system to perform its intended function.

I have actually been using one of these drives in my laptop for about a week now and I do think that my system is a bit faster as a result although I have not quantified performance.  Other approaches will use separate HDDs on the SATA interface and solid state storage that may be separate SATA (or mSATA) devices or may use other interfaces such as PCIe.

My belief is that this will be a better path for wide-spread adoption of solid state storage in computers where they improve overall system performance with some flash but not enough flash memory to attempt to replace a HDD; the economics of that are just not there.

El Reg: Will single-platter 2.5-inch hard drives be successful? How much cheaper are they to make than standard dual-platter 2.5-inch drives? Could we see hybrid single platter drives, for example in tablets where they could provide near-flash performance at a lower price than flash.

No reason why not, we have had 1-platter 3.5-inch and I think 2.5 inch drives for a while.  By not having a second disk and two heads I think HDD companies will save $10 to $15 on their bill of materials costs and thus could either offer a lower price 2.5-inch straight HDD or perhaps as you suggest a low cost hybrid drive.  The trade-off is lower storage capacity with 2 disk surfaces instead of 4 (about a 2X factor).  An interesting aspect of a single platter 2.5 inch drive is that it could be thinner than a 2-platter disk drive and thus fit better into thin tablets and other products.

Disk drive capacity increase challenge

El Reg: How do you see the HDD industry solving the challenges of increasing disk drive capacity facing it as current PMR technology runs out of steam? Do you see the transitions to the next technology being (a) Shingled writing or adding platters or both while staying with PMR (Perpendicular Magnetic Recording), (b) then HAMR (Heat-assisted Magnetic Recording) or a similar thermally-assisted technology, (c) then DIscrete Track Media or Bit-Patterned Media or both?

Adding platter is the easiest but it also adds overall cost and if too many platters, drive thickness so this only gets you so far.  Shingle writing requires work on the servo system and track following (so firmware and electronic issues) but it requires no changes in the media and little if any on the heads.  However there are significant architectural and use case issues.

Write performance could suffer once the drive is fully written since a rewrite process would be required to add new data.  The memory management would get a bit like flash memory with their erase/write cycles. In addition to the performance hit the increase in TPI (tracks per inch) is probably only 30 per cent and at most 50 per cent and is a one time only bump.  As a consequence by 2014-2015 we will need some thing new such as HAMR or patterned media to continue the areal density growth.

These technologies still have significant technology or capital cost issues that get in the way of implementation so we may see a temporary decline in areal density growth to maybe 20 per cent for a couple of years.  Of course breakthroughs happen and like in semiconductor line width we could see paths to continued areal density increase that avoid a slow-down.

El Reg: What are the timescales and areal densities and capacities involved with each of these?

Unaided perpendicular recording gets us to about 1Tbit/in2, shingle writing gets us at best to 1.5Tbit/in2 (and by the way WORM type applications such as backup and DVRs may do fine with this approach).  If we are to achieve 2Tbit/in2 or higher we need some new breakthrough that is currently unclear or HAMR or patterned media.

El Reg: Will we see 12Gbit/s SAS interfaces? Will drives with additional platters need such a faster interface?

I think faster interfaces are quite likely.  In particular SSDs can already saturate 6Gbit/s interfaces today and could also do that tomorrow. Also these channels (or controllers) may be shared between devices and thus we need a higher composite data rate to support these devices than any one device would require.

El Reg: Will we see pure USB interfaces for external drives with no SATA interface at all and no need for a SATA-to-USB bridge inside the drive?

Maybe, we could see more specialisation in drives for external storage such as this, combined with shingled writing as well perhaps?  Also what about Light Peak?

Coughlin is seeing two intermediate technologies capable of boosting capacity before there is a transition to post-PMR Technologies. These are adding platters and heads on the one hand, and shingled writing on the other. My opinion is that we could see these in 2011 and 2012, These would be followed by either HAMR or patterned media. These I reckon are a post-2013 story.

It appears that the use of single platter drives and hybrid SSD-HDD products would enable HDD technology to withstand flash replacement of HDDs in some sectors of the HDD market. These will be a 2011 story, possibly a late-2010 one as Seagate's Momentus XT is out there already. ®