In March of last year I began writing about an IBM research effort to develop a new hardware and software architecture for ingesting and analysing large scale streams of data in real time; what they call stream computing (here, here, and in HPCwire here). Last week IBM announced some of the results of that effort
Today, IBM unveiled a revolutionary prototype of the world’s fastest automated options trading system. During the project, scientists at IBM Research collaborated with TD Securities to achieve a 21 times performance improvement on the volume of data consumed by financial trading systems.
…Through the combination of IBM’s InfoSphere Streams — a breakthrough software technology from IBM Research — and IBM’s Blue Gene/P supercomputer, the IBM Research team created a unique stream processing system ideally suited to meet and surpass the demands of the financial services industry. By enabling rapid, intelligent analysis of live streaming data from a practically unlimited number of sources, IBM delivered astoundingly low latency — the time between when data is received and when it’s acted upon — far surpassing the performance of traditional trading systems.
More in the news release.
I’m very interested in this topic but I find the article really vague. What are they able to do so much faster? Are they able to calculate correlations instantly, find arbitrage opportunities, calculate VaR on the full portfolio, … ? And I’m unable to find any papers on this either?
Toon – they haven’t talked much about it. When I talked with IBM last year they were publishing a little, but being cautious about it while they figured out what the business is. The HPCwire article I wrote last year has general info about the approach, but obviously not this specific application. At the time TD wouldn’t talk in other than general terms because they view this stuff as competitive advantage. I’ll bet the same is true today, which probably explains the vagueness.
John, thanks for the comment. I remember the last article and at the time I also contacted the guy listed on the IBM website but heard nothing back. Also last time IBM/TD anounced they could do something great without specifying what exactly.
In finance more and more companies are talking about supercomputing but very little is published much in contrast to e.g. the Computer-Aided-Engineering world. If this is to defend the ‘competitive advantage’, then how come SANDIA is nevertheless publishing a lot !?
Toon, in case you are interested, here is an article in MIT Technology Review. http://www.technologyreview.com/computing/22463/?a=f
Not much there either that isn’t covered in the other articles I’ve pointed to.
Is this connected to the Exascale Stream project IBM has in Ireland: http://www.ibm.com/news/ie/en/2008/11/25/e502955h74656w51.html
You Made Some Good Points There. I Did a Search on the Topic and Found Most People Will Agree with Your Words. Thank You!
P.S: Please Take a Minute to Visit My Website as Well: http://snurl.com/stockassault