Edit

SeaMicro

http://www.seamicro.com/

Servers today consume more than 2.5% of the total electricity used in the United States—more than $4 Billion dollars each year. Over the last six years, the power consumed by servers has more than doubled.

For companies in the data center, power and space combine to account for more than 70% of operating expense. In fact, research has shown that over a server’s lifetime, the cost of power exceeds its purchase price.

SeaMicro has identified the primary drivers underlying power inefficiencies in servers and has systematically rectified these shortcomings. The result is a small form factor server that uses one-quarter of the power and requires only one-sixth of the space used by traditional servers. SeaMicro servers are plug-and-play—they require no changes to software operating systems, applications, or management infrastructure.

Over the past 10 years, the data center has undergone a sea change in size and scope, including dramatic changes in the demand for compute, the type of compute required, and the economics of operation. Scale out has replaced scale up, and the Cloud has become the home of many business services and a source of on-demand compute. Despite these changes, the server remained architecturally unchanged. Server manufacturers made no accommodation for the new and different workloads and traffic patterns despite the highly specialized workloads that rose to dominance in the data center. This mismatch between specialized workloads and generalist servers is an underlying cause of the power consumption challenge in the data center.

SeaMicro identified this mismatch as an opportunity, but unlike previous approaches to reducing power draw in servers, did not focus on redesigning power supplies or improving airflow. Instead, SeaMicro brought together technical insights from CPU design, virtualization, supercomputing, and networking to create a new server architecture optimized for scale-out infrastructures. SeaMicro characterized the specific computational requirements of scale-out work in data centers and designed custom hardware, ASICs, and FPGAs, and selected CPUs optimized for scale-out workloads to create the industry's first purpose-built server optimized for workloads found in the hyperscale data center.