A New Way to Use SSDs
In his Flash Memory Summit keynote on Wednesday, Micron VP and Chief Memory Systems Architect Ed Doller made a compelling demonstration of the power and performance advantages of a new approach to computing.
With true showmanship, Doller had his co-workers hand out buttons with LED lights to the entire audience. The LEDs in these buttons were either green or blue, with the colors randomly dispersed among the crowd. Doller asked the entire audience to turn on their lights, then called one row of the audience to file up to the stage so he could determine whether each person’s button was blue or green.
He pointed out that this was like having a single CPU check the contents of a drive. He then asked why things should work this way – wouldn’t it be more sensible to ask the audience members to determine the color of their own buttons without having to go up to the stage for him to look them over?
Doller then told us about the results of an experiment Micron tried in which the ARM processors within standard Micron M500 2.5″ SSDs were loaded with a modified version of a MySQL database search program. Micron found that the performance of searches improved linearly with the addition of SSDs, and that the burden on the server’s processor dropped from 100% to just 1% as the server passed the responsibility of performing the search to the SSDs.
The new system, using eight SSDs, performed twice as fast as the conventional system while consuming only 61% of the power. With 32 drives the system’s performance rose to eight times that of the conventional approach, using only 16% as much power.
Doller created a new term for this approach. In a world of “Scale Up” vs. “Scale Out” he decided that this should be called: “Scale In.”
What enables the higher performance, lower power, and decreased server utilization is the fact that we now have pervasive computing. There isn’t just one CPU in any system, there are numerous CPUs scattered throughout. Why not free the unused resources of these multiple CPUs to act on the data near where the data actually resides?
While Doller didn’t make it clear whether Micron intended to ship SSDs that were documented to support outside development of such software, he illustrated an important point – we are at a juncture at which major systems are likely to stop using today’s common centralized server model to one in which smaller distributed processors may be used closer to the data to achieve significant performance and economic benefits.