– Andrew Feldman, CEO at SeaMicro (w ww.seamicro.com), says:
Have data center servers changed much in recent years?
The obvious question is why has there been so little innovation in the server space in the last 15 years.
Our view is that server companies became supply chain companies, focused on wringing every last ounce of cost out of their supply chain. They moved to low cost manufacturing regions, used massive purchasing power to pressure component vendors etc.
What they didn’t do was invest in R&D or invent new technology. They didn’t innovate. Even the so-called ‘blade revolution’ is largely a packaging exercise that takes a 1 RU server and rotates it 90 degrees and slides it into a backplane. Instead of real innovation, server companies relied on the next Intel reference board, the next processor…
SeaMicro’s approach is very different. We invested tens of millions of dollars in research and development. We invented technology that allows us to remove 90 percent of the components from the mother shrinking the motherboard to the size of a business card. We then developed technology that could link hundreds of these mini-motherboards together at radically reduced cost and power draw.
And the result is a new server architecture that uses one-quarter the power and takes one-quarter the space.
Today, most of the server manufacturers are focused on modifying existing designs.
They are working to remove a component here or there. At SeaMicro, we don’t think that is a productive approach. Our view was that the server needed to be rethought from the ground up. The question isn’t ‘what component should we change’ but rather, if we were to begin with a blank sheet of paper, which components would we use, and what technologies would we invent to eliminate dozens of components?
The most important trends are around the rise of the Internet workload and the recognition that one server size does not fit all needs.