Server Innovation May Boost Search Performance
June 16, 2010
The goslings and I like hardware, but we focus on the antics of the search and content processing sector. Truth be told we would rather build a gizmo than write about how to filter Chatroulette results. We want to highlight what we think is one of those hardware innovations that could open the door to more sophisticated content processing. Instead of waiting for a quantum computer to crunch Big Data, you will be able to buy A SeaMicro SM10000 FOR $140,000 and get your hands on a low power, 512 Atom processor fire breather.
You can get a quite good summary of the SeaMicro innovations in “SeaMicro Drops an Atom Bomb on the Server Industry.” Let me highlight three points from the write up and offer an observation from a flight departure lounge, not my usual goose pond.
First, the SeaMicro approach is to use Intel’s Atom chip, a low power, small gizmo. With proprietary hardware and software, SeaMicro chops the power requirement so that high performance comes with lower energy consumption. The cost of power for data centers is now a deal breaker. Electricity bills can add up to more than the cost of the server hardware in 36 months or less.
Second, the SeaMicro server uses some clever virtualization methods to make small lower power processors work like their larger, more inefficient big brothers. A SeaMicro server plugs in and operates like a data center in a box. Hook several SeaMicro servers together and you get the type of horsepower that AT&T has embedded in its most sophisticated infrastructure. Imagine. An Ashburn an a Dallas in your local hosting facility.
Third, the gizmos are plug and play. Remember the exciting days of Sun Microsystems system certification? Gone. Total cost of ownership on the SeaMicro servers is reduced. The article says “by 75 percent”. Half that would make me turn cart wheels.
What are the implications?
First, content processing bottlenecks could become less of a problem if the information retrieval system runs on a SeaMicro server. Search and retrieval is a weird combination of hurry up and wait with nasty demand spikes. Why wait for a software vendor to improve their software. Throw SeaMicro methods at the problem.
Second, the company has rolled out an innovation that I thought would come from the companies with the big, old fashioned data centers. I expect some scrambling by those caught flat footed by SeaMicro.
Finally, those in research computing will want to get their hands on one of the SeaMicro gizmos. With more fire power, ideas that were impractical on older servers may become doable.
SeaMicro may not end up the big winner, but its innovations warrant some attention.
Stephen E Arnold, June 16, 2010
Freebie
Comments
-
- Subscribe to Beyond Search
Feature archive
News archive
-
