Seven Barriers to Server Consolidation
We set out to find what is holding back server consolidation, in the hope that identifying the problems will spur a stronger drive to overcome them - and produce a leaner, meaner data center in the end.
Low-power "micro" servers may be the biggest thing in hardware development circles, but can they really handle the kinds of data loads that confront enterprises these days? And can they perform at such a level and still produce a net reduction in energy consumption?
Word from the vendor community is yes. In fact, low-power processors like the ARM and Intel's Atom might be even more adept at handling Big Data than the mobile applications that have been their stock and trade. Caldexa CEO Barry Evans explained it to GigaOM's Derrick Harris this way:
Big data is a great fit for us and ARM servers for three key reasons. First, it is an inherently scale-out application, requiring a lot of efficient processors. Second, it is a fast-growing market place without a lot of requirements for legacy baggage. Third, the application software is widely available to run on ARM today.
To further ARM's reach into the application environments in use at most enterprises today, Caldexa recently founded the Trailblazer Initiative with the intent on devising an ARM ecosystem that would cater to cloud computing and other enterprise-class endeavors. The company has lined up a number of software partners, including Autonomic Resources, Canonical, Gluster and Opscode, which will gain access to Caldexa's hardware, operating software and toolsets to help foster an integrated ARM environment.
To make any real headway, however, Caldexa and others will have to displace Intel as the leading supplier of enterprise silicon. And that's a tall order, considering the company is already planning for the end of multi-socket dominance in favor of single-socket micros. The Inquirer's Lawrence Latif even wormed out of Dylan Larson, director of the Xeon platform, that Intel's approach will likely be some form of suped up Atom released under the Xeon moniker - quite possibly a multicore design with advanced features like ECC memory support.
Core processing is not the only place where the low-power battle is being fought, however. Rambus, for example, recently unveiled a new clocking technology aimed at reducing memory subsystem requirements in both mobile and compute environments. The system features a "feed-forward architecture" that enables memory to jump from idle to active in only a few nanoseconds without the usual circuitry found in current memory systems. The company says it can keep the power envelope below 2.4 mW/Gbps on a 40 nm CMOS process.
In today's economic climate, it's hard to argue against anything that is low power. However, the fact remains that these new architectures are largely untested in real-world enterprise environments. There's no reason to think they can't handle the pressure, but it's probably best to wait and see what they can really do before popping the champagne cork.