9TB in 20 minutes? Sign me up!
IBM hails 'breakthrough' algorithm
In an experiment IBM researchers used the fourth most powerful supercomputer in the world - a Blue Gene/P system at the Forschungszentrum Julich in Germany - to validate nine terabytes of data in less than 20 minutes, without compromising accuracy. Ordinarily, using the same system, this would take more than a day. Additionally, the process used just one percent of the energy that would typically be required.
In a press release last week, IBM hailed a "breakthrough method based on a mathematical algorithm that reduces the computational complexity, costs, and energy usage for analyzing the quality of massive amounts of data by two orders of magnitude. This new method will greatly help enterprises extract and use the data more quickly and efficiently to develop more accurate and predictive models."
OK, I’m a sucker for anything that does something a lot faster – even if I don’t quite understand how it does it.
So I have to write a least a little bit about the paper that three IBM researchers submitted documenting their technique and results. I was faced with a blizzard of phrases such as “Inverse covariance matrices,” “Matrix factorizations,” “Cubit cost,” and this helpful explanation: “First, we turned to stochastic estimation of the diagonal.”
These terms, plus many others that I also don’t understand, are in just the abstract; the body of the paper seems considerably more technical and complex.
The one phrase that I fully understood was: “We stress that the techniques presented in this work are quite general and applicable to several other important applications.” And this is an important phrase, because it means that using this technique (and others that smart guys are working on right now), we’ll be able to see orders of magnitude improvement in other analytic tasks that use vast amounts of data. It’s always good to see progress.
In case you’re interested, here are some pictures of the guys who came up with it – mostly shots of them standing around looking smarter than any of us.
Sponsored: Hyper-scale data management