Will ARM Face a Cold Shoulder in the Data Center?

 

By Charles King, Pund-IT, Inc.  January 8, 2014

Over the past couple of years, there has been rising interest in and considerable promotion of servers and appliances based on the ARM microprocessor architecture. At first glance, the narrative makes some sense. ARM’s native performance and energy efficiency, when multiplied across hundreds or thousands of systems, looks effective for supporting certain kinds of workloads common in Web 2.0, social networking and search sites. That is especially the case when ARM silicon is leveraged in concert with innovative fabric technologies, like the Freedom Fabric solutions that AMD acquired when it purchased SeaMicro in 2012.

As a result, ARM has generated a good deal of heat among venture capitalists and other investors; at last count, over a dozen start-ups are developing ARM-based server silicon. But just before Christmas, one of the best known of these players—Calxeda—went belly-up and quickly shut down operations. Industry finger pointers were fast on the draw claiming, with enviable 20/20 hindsight, that the company had been mismanaged, that it had a poor business plan and that it wasted cash resources developing solutions based on existing 32-bit designs when it should have bided its time until 64-bit ARM was available.

These explanations were practical enough but we think the challenge faced by Calxeda (as well as other ARM server players) is actually a bit simpler; energy efficiency has long been a tough primary selling point in the global technology market. That’s partly because such solutions typically require customers to make stiff up-front capex investments in order to capture longer term opex savings. In addition, while “green” IT may feel good, perceptions of its value vary widely due to energy availability, fluctuating power costs and local business imperatives.

That’s certainly been the case in similar past efforts, such as IBM’s “Project Big Green” initiative. Launched in 2007, Project Big Green aimed to use new and existing tools to maximize the energy efficiency of enterprise data centers. Looking back, IBM was prescient in leveraging technologies and strategies that eventually became common currency across the industry, including robust consolidation through virtualization and granular measurement of energy usage.

But while the company itself gained considerable financial benefits by deploying these solutions in its own massive data center infrastructure, the effort never achieved the sort of commercial success the company hoped. That was partially due to timing. Big Green arrived toward the end of the energy crisis of 2004-2008 as oil costs peaked and then began to decline making efficient power usage appear less compelling. The situation was further complicated by the 2007-2009 recession when even energy enthusiasts were reducing or halting IT investments.

The financial environment is obviously different today than it was then, but we believe that some recent events are likely to negatively impact the health of energy efficient IT solutions. For example, the ongoing boom in natural gas production in the U.S. has resulted in an abundant supply and dropping costs. More importantly, though, utilities across the country have seized on the boom to justify retrofitting aging coal burning power plants for natural gas, thus gaining efficiency and producing electrical power which is considerably “greener” than ever before.

That’s enough to throw ARM’s value proposition off kilter. But then toss in the general unfamiliarity of new system architectures (implying retraining for data center and maintenance staff), retooling of management processes, the need to invest in new ARM-compatible applications and continually constricting IT budgets, and ARM’s benefits begin to pale. Finally, consider the innovative work that Intel, AMD, Dell, IBM, VCE and others are doing to improve the efficiency of fully-established server architectures, and the opportunities for ARM begin to look positively ghostlike,

That doesn’t mean that these solutions are doomed by any means. For businesses with considerable in-house IT talent and massive hyper-scale data center infrastructures, like Google, Facebook and Yahoo, ARM has been and is likely to remain of interest. They are, after all, the primary targets that ARM players have been aiming for all along. But we believe that Calxeda’s sad demise and broader market trends and events have measurably dimmed the likelihood that ARM will disrupt the IT infrastructure marketplace to the degree that proponents have hoped.

© 2014 Pund-IT, Inc. All rights reserved.