When AMD general manager Andrew Feldman claimed at the Open Compute Forum at the end of January that ARM-based microprocessors could take as much as 25 per cent of the server market by 2019, it caused more than a few raised eyebrows.
The reason? Apart from a bundle of Sparc microprocessors from Sun and the Power architecture from IBM, the data-centre market is completely dominated by Intel with a share of as much as 95 per cent.
Yet the SeaMicro co-founder was suggesting that in just five years lower-cost alternatives from companies in the ARM eco-system could take a huge chunk out of Intel's de facto monopoly.
Feldman argued that while data centres are booming - both in number and size - their role is changing, and those changes are playing to ARM's architectural and eco-system's strengths.
First, smartphones and ubiquitous mobile internet access "has not only changed the way that we interact with the world, but it's put the data centre at the centre of our daily experiences", says Feldman. Almost everything that someone does on a smartphone or tablet ends up, in some way, in the cloud and, hence, in a data centre.
That has led not just to an explosion in the number of data centres, but also their size and what they are doing. It has shifted the epicentre of "big IT" from the corporate world and is concentrating it in the hands of a number of major companies, such as Amazon Web Services, Google, Facebook, Salesforce.com, NetSuite, Rackspace and myriad other big data-centre operators.
That means that microprocessors that use, say, 10 or 20 watts of power instead of 60 watts, will run very much cooler, so that many more of them can be crammed into the same space, while also enabling the operator to offer more compute for the same electrical power output - an increasingly important consideration for many large data centre operators constrained by the capacity of local power grids.
On top of that, ARM microprocessors are typically much cheaper than Intel's and while Intel is the world's biggest, most profitable and most successful semiconductor company, some forces are immutable, claims Feldman. "One of the things we've learned in the history of compute is that smaller, lower-cost, high-volume CPUs have always won - without exception," he says.
Furthermore, only one-thirds of the world's population are currently connected to the internet in any way - but the other two-thirds will connect very soon. "We've got 6.1 billion cell phone subscriptions but only one billion smartphones. So we know the internet is going to be delivered to these people by a phone. We know how to get to them and we know this will put tremendous pressure on the data centre," says Feldman.
Hence, the trends that have driven data centre growth across the developed world will continue for the foreseeable future.
[Turn to next page]
By eliminating high entry costs for big data analysis, you can convert more raw data into valuable business insight.
A discussion of the "risk perception gap", its implications and how it can be closed