Mark Palmer
Sun, 26 Aug 2007 22:22:05 -0500
James Kobielus with Network World wrote a recent article about CEP called: "
Complex Event Processing: Still on the Launching Pad." The author seems to equate being "off the launching pad" to being adopted by
business intelligence vendors. The article made me wonder why anyone would equate success of a new technology by it's adoption and usage in a style it's designed to improve. It's like saying that the internet will be successful once you flop down on your couch and watch the evening news on a web browser.
CEP's strongest use is for computer-based automation; not human decision support. CEP's success should not be measured by it's use as human decision support. CEP often
replaces manual human decision making - as a result, it empowers a new class of knowledge worker and application metaphor.
Take
algorithmic trading, for example. Trading is a business "decision making" function that has been performed manually for years by MBA's with a flair for identifying opportunities and acting quickly. Traders apply the highest-end "business intelligence" there is. CEP is being used to automate trading because much of what these MBAs used to do can be automated. The head of FX trading at a large investment bank I recently visited, for example, told me that in the past 4 years their trading volume has increased by over 10X; at the same time, their trading staff has been reduced by 80%. Mind you, traders are not going away. Their role is changing from the old days of relationship-based, click-based trading to more of mission-control responsibility, monitoring and optimizing algorithms as market conditions change. Similarly, CEP is being used to
automate fraud detection. Instead of having loads of surveillance staff visually watching for suspicious patterns of activity, a computer can monitor activity for fraudulent patterns of behavior, and alert staff when suspicious activity is detected - in real-time.
And CEP can help do these kinds of automated systems more quickly and at greater scale. Even the best traders, for example, can only monitor between 5 and 10 "spreads" (the difference in price between two instruements) in their head at a time. A CEP engine can monitor hundreds of them. Decision can be made and executed in less time than it takes a neuron to fire in the human brain, and for the trader's finger to be moved to press the "buy" button (about 80 milliseconds). Finally, CEP can be used to build these applications more quickly than backward-looking BI programming styles, because CEP contains built-in notion of time, and the built-in ability to detect patterns of events (A followed by B, then C). BI isn't designed for any of this.
So industry observers should watch for applications that
replace large banks of analysts using BI tools; not for BI vendors cannibalizing their own market with new technology that supplants the old.
Source...
