Original URL: https://www.theregister.com/2014/11/10/kryders_law_of_ever_cheaper_storage_disproven/

Kryder's law craps out: Race to uber-cheap storage is voer

Bad news for cloud and archive

By Chris Mellor

Posted in Storage, 10th November 2014 15:03 GMT

Analysis Kryder's Law says disk storage will keep on getting cheaper. Yet it is increasingly looking like it won't, which has baleful implications for cloud storage and archiving.

Like Moore's Law, Kryder's Law is (a) not a law, merely a wonderfully apt observation for a prolonged but not eternal period of time, and (b) borne of an incredibly bold and confident view of technology.

Mark Kryder was a Seagate SVP for research and its chief technology officer when an article titled Kryder's Law – where he was interviewed – appeared in a 2005 issue of Scientific American. The "Law" declares that disk drive areal density would more than double every two years, meaning disk drive capacity would do likewise.

The article said:

Since the introduction of the disk drive in 1956, the density of information it can record has swelled from a paltry 2,000 bits to 100 billion bits (gigabits), all crowded in the small space of a square inch. That represents a 50-million-fold increase.

That 50 million increase took place over 41 years, a rate of improvement fantastically faster than Moore's Law, which, the article said involves, "doubling the power and memory of computer semiconductors every 18 months".

The SA article got it spectacularly wrong about small hard drives replacing flash drives when it said: "Soon hard drives will migrate into phones, still cameras, PDAs, cars and everyday appliances," and replace flash drives. Exactly the reverse happened.

Kryder wrote in a 2009 PhysOrg.com article (a précis can be found here) that, by 2020, a 2.5-inch, 2-platter drive would store more than 40TB of data.

Currently 2.5-inch drives are at 500GB/platter with some at 600GB or even 667GB/platter – a long way from 20TB/platter. To reach 20TB by 2020, the 500GB/platter drives will have to increase areal density 44 times in six years. It isn't going to happen.

What has happened is that it is getting progressively more difficult and much more expensive to keep on cramming bits onto disk platters. Kryder's Law is "slowing down", so to speak.

Rosenthal's take

David Rosenthal is an engineer who worked with James Gosling at Carnegie-Mellon University on the interface in the Andrew Project, moving on to operating systems' windowing interfaces with Gosling at Sun. He then moved to Nvidia, as employee number 4, to work on graphics chip. His DSHR blog focuses on long-term data storage.

In May this year, he blogged about a talk he gave at Seagate entitled "Storage will be much less free than it used to be." His speech reviewed the progress implied by Kryder's Law:

After more than a quarter century of Kryder's Law, the exponential increase in disk capacity at constant form factor, and thus exponential decrease in $/GB, almost everyone believed that long-term storage was effectively free.
Disk capacity increase

Kryder's Law: disk capacity increase trend 1980 - 2010. Source David Rosenthal.

The assumption was that: "If you could afford to store some data for a few years, you could afford to store it forever. The cost of that much storage would have become negligible." Only that is no longer true.

Rosenthal discusses "the Kryder rate (the annual percentage drop in $/GB)" and says it is slowing, citing a graph from Preeti Gupta at UCSC in support.

Kryder slowdown

Kryder slowdown. Chart by Preeti Gupta at UCSC.

This graph "shows that disk is now about seven times as expensive as it would have been had the industry maintained its pre-2010 Kryder rate."

Kryder rate slow down

The 2011 Thai floods almost doubled disk capacity cost/GB for a while. Rosenthal writes: "The technical difficulties of migrating from PMR to HAMR, meant that already in 2010 the Kryder rate had slowed significantly and was not expected to return to its trend in the near future. The floods reinforced this."

HAMR is heat-assisted magnetic recording and the costs of migrating to this from current PMR (perpendicular magnetic recording) are anticipated to be huge. Rosenthal points out that "[i]n the real world, exponential growth can't go on forever, it is always just the steep part of an S-curve. It is now widely accepted that Moore's Law has slowed. Although transistors will continue to shrink for a few more generations, this will no longer result in transistors getting cheaper or CPUs getting faster."

Here's a key insight:

... what looks like a smooth Kryder's Law curve is actually the superposition of a series of S-curves, one for each successive technology generation. Naturally, because the easy transitions get done first, the cost of each successive transition increases, perhaps even exponentially. Since margins are constrained and so, these days, are volumes, to generate a return on the investment in each transition requires that the technology be kept in the market longer. The longer interval between transitions translates to a lower Kryder rate.
Kryder esses

Kryder esses in a graph by Dave Anderson via David Rosenthal.

This slowing Kryder rate means that the service life for disk drives will lengthen. Each step change increase in capacity will take longer, be relatively more expensive, and disks will have to be engineered for longer life, meaning lengthier warranty periods.

Rosenthal also works out the effect of a 20 per cent per year Kryder rate (from IHS iSuppli), IT budget growth at two per cent/year, an IDC data growth projection at 40 per cent/year and storage at 12 per cent of the IT budget.

He reckons that, after eight years of this, the annual cost of storing all the data accumulated since year zero, relative to the cost in year zero, is 100 per cent of your IT budget.

Race to zero

Let's turn to the cloud and the Amazon/Google/Microsoft cloud storage cost reductions. The three are in a race to the bottom, a race to zero, so to speak, as they try to wipe out other cloud storage competitors and persuade corporates to store their data in their clouds.

Aaron Levie, the CEO of file sync 'n' share startup Box, which has raised $559m, has said: "We see a future where storage is free and infinite."

He is quite, quite wrong. It won't be free and it won't be infinite. And it won't be quite so necessary to provide added-value compute and other services alongside cloud storage to attract saying customers. Storage will change from its anticipated positioning as a loss leader. In the longer term it won't be a loss leader and suppliers will be able to charge for it.

The Kryder rate slowdown will see to that and business models based on a continuation of the Kryder at at pre-Thai flood levels will fail as their costs become untenable.

Already some infinite storage in the cloud models, like Bitcasa's, are being ended. The storage times they are a-changing and Kryder should soon find his law has crapped out. ®