vincent
09-18-2008 07:50 PM
Here are some of the highlights from this week’s Gartner Event Processing Summit:
- Roy Shulze commented that there were 3 main areas of Complex Event Processing specialization: (event) capture, computation, and response. A possible 4th was (event / EP) management. Also interesting was his prediction that “in ~5 years, the market would move from tools to applications” - probably depending on your meaning of “application”. TIBCO BusinessEvents was classed as “Ultra Low Latency CEP”, with TIBCO Spotfire getting a mention for retrospective analytics.
- Mary Knox covered EP in the Banking and Capital Markets (or whats left of them), with some interesting statistics. For example, 37% of financial services organizations today consider themselves using “interaction SOA” and EDA, with 6% using (dedicated?) EDA. In terms of IT investments in CEP and BAM, Gartner see investment services companies’ annual investment going from $173M to $355M in the US (and $161M to $342M in Europe) from 2008 to 2011 (albeit including services etc). Solutions will focus on things like fraud (detection), Quality Of Service, exception detection and processing, market opportunity and risk, algo trading, and credit / system status monitoring.
- HG Trading showed a small-company use of event stream processing to do pairs-trading (in other words, their complex event was made from comparing 2 events). Although this might not be a technical “big deal”, the use of a CEP technology to benefit a small company* certainly shows that CEP is not just for large enterprises.
- Marc Adler of Citi certainly didn’t seem to do his “chosen vendor” any favors with his “forthright” analysis of dealing with the CEP industry. He mentioned his wish-list of out-of-the-box features [*1] being: KDB support,* FIX adapters, SQLServer as an event source, TIBCO EMS support, full 64′ .NET3.5 integration, analytic tools, integration with things like TIBCO Spotfire, …. In particular he noted some of the perils of dealing with smaller vendors, like having to train them in his investment banking domain [*2]. Surprisingly, he also complained about the lack of flexibility in the streaming SQL approach to CEP he had chosen, such as a lack of granularity / difficulty in doing simple things. Hopefully he won’t waste too much time trying to solve his CEP problems with just streaming SQL. But overall, the presentation was very well received, and probably the best of the show.
- Arno Jacobsen of Toronto University’s middleware team gave a progress report on his R&D project using event monitoring for BPM SLAs. I thought it interesting he was advocating dynamic (on the fly) pub-sub subscriptions to “simplify development”, although of course in the real world of business-critical messaging systems, your IT department might not be too happy about you doing development on the production network…
- David Luckham made some good observations, like the need for business event hierarchy modeling, and the onset of “holistic event processing” (covering a community of evolving and interacting CEP applications) for things like Earth Catastrophe Monitoring, International Epidemiology Monitoring, etc.
:
[1] Easily do-able with TIBCO BusinessEvents, in case any OEM is interested in a custom solution for the investment market…
[2] Note that vendors tend to have product AND industry specialists. The latter don’t come cheap, so this tends to be why smaller vendors sometimes either specialize in vertical markets or OEMs.
Source...
09-19-2008
