Who will build the next generation of physical cloud infrastructure?

Big computer manufacturers are leaving their traditional markets to reposition themselves as cloud providers. At least one global name is entering the field they're leaving, but will it be enough to build the infrastructure that drives the cloud in the future?
Written by Drew Turney, Contributor on

According to IBM's 2013 annual report, the company's cloud business grew 69 percent last year, delivering US$4.4 billion of revenue. At the same time, slow hardware sales have seen revenue dip for the last seven quarters.

Early last month Hewlett Packard said it would invest US$1bn in cloud computing over the next two years, a move the company says is necessary to ward off competition even if it eats into its own hardware business. HP is also moving its high volume server production to Chinese manufacturer Foxconn.

AT&T is also moving from hardware to software, revenues from which saw a 15 percent year-on-year growth in 2013 and now represent $9 billion a year — more than a quarter of the company's income.

The above companies all made their names in manufacturing or selling computer or telecom equipment and infrastructure, now they're all easing away from their former bread and butter in favour of the software that will run it all.

But someone will still have to build servers and computers — more than ever before as data centres spring up worldwide at an ever-growing rate. But as the name brands move strategies away from hardware, who will that be?

As Ben Blair, co-founder and CTO of online hardware exchange MarkITx, says: "In 2008 HP, Dell and IBM bought 75 percent of Intel's chips. By 2012 their activity was much lower and Google had moved up to the number five spot."

A new hardware class

Some think the cloud will be stored and driven on white box technology – no-name computers and datacenter processors assembled for peanuts in their millions using mass-produced, non-specialist parts.

Others say initiatives like Facebook's Open Compute are the future. The company aims to create a wide-scale, open, bare bones computing infrastructure by re-engineering everything from network architectures to motherboards and power supplies to find new, low cost efficiencies and alternatives.

Software and service integration will be very much the new black if the world keeps moving in that direction. The cores, racks and boxes that make up the physical cloud infrastructure will cost an ever-decreasing per-unit amount, all as dispensable and hot-swappable as a memory card in your digital camera.

It will also mean a whole middle class of cloud providers might not be able to compete with the gains that will become available to volume players thanks to cheaper hardware and networks.

"If smaller providers can't differentiate themselves from the big players they're going to move to Open Compute to stay in the game," says MicroStrategy Cloud vice president and general manager Ian Matteson.

That said, big cloud operators might not be making as much as we think from bare-bones IaaS. Despite dramatic price drops, Martin van Ryswyk, engineering vice president at Apache Cassandra distributor DataStax, says the industry giants are using more white box hardware too.

"Even big cloud operators are relying more on software to scale services rather than robust hardware."

The box isn't dead

But before we imagine a utopian future of egalitarian hardware processing our hopes and dreams of ubiquitous cloud computing, there are other factors to consider.

Firstly, not everybody is 'dumbing' boxes and disks down to such a degree, and not every hardware manufacturer is having a hard time. Oracle's latest earnings announcement revealed steady growth in revenues from its Engineered Systems business, which accounts for nearly a third of total hardware sales.  The Engineered Systems are pre-integrated "stacks" (including software) and designed for specific purposes, like database, compute, or business intelligence. That's a long way away from commodity boxes.

Disk drive developer HGST Hardware is another example. Nigel Edwards, sales and channel marketing, Europe for HGST, has a futuristic-sounding helium-filled drive.

"Traditional air-filled five-platter hard disks can't provide the level of capacity customers are demanding," he says.

The only way to increase the capacity inside a disk, Edwards explains, is to increase the number of platters and heads. Helium allows for each platter to be thinner, is less prone to drag and vibration from air resistance, there are fewer humidity changes or contaminants, and in the end yields 50 percent more capacity from 23 percent fewer watts of electricity.

Even if the move to a no-name physical cloud infrastructure does happen, it'll take time.

Charles King, principal analyst at IT consultancy Pund-IT, says interest in public cloud computing is taking longer to pan out than the original optimism indicated.

"Major cheerleaders are now conceding that organisations will maintain numerous processes and vast amounts of data in their own firewalled datacentres and private clouds," he says.

King adds that a white box infrastructure world is only suited to major players who deal in volumes most of the industry can only dream of.

"Original design manufacturer (ODM) hardware is attractive in high volumes where you buy thousands or tens of thousands of systems, not hundreds or a few dozen. Google, Facebook, Microsoft and Yahoo certainly fit that description but the vast majority of organisations don't."

The cloud and you

In the wash-up, what difference will it make at the coalface of your day-to-day business activity? Same as always — very little, as long as it works.

"It's the service layer that carries the most weight in a cloud world," says Robert Miggins, business development senior vice president for Peer 1 Hosting. "That's why IT decision makers are going straight to the service providers for their needs."

Still, the advances in the underlying physical infrastructure will drive its uptake, and in turn, evolution. MicroStrategy's Matteson reminds us that custom IaaS providers are using the big providers' services at the same time they're directly competing with them. While essentially offering the same hardware, they're at a disadvantage.

"Costs are mostly the same across the board," he says, "it's the labour related to manual deployment that drives the price up."

No matter who you pay for cloud services, the value proposition will be the service, not the machinery in the back room. The only difference now is that the former big hardware makers are jostling to get some of your cloud service budget too, and as they concentrate on customer-facing technologies, they might leave a vast, untapped market behind in their wake.

Editorial standards