Two of many interesting tidbits of news from last weeks internal TIBCO meetings were:
- TIBCO competed against a C/C++-based CEP product normally seen in Financial Services Front Office applications (i.e. extreme low latency applications). But because TIBCO BusinessEvents could effectively exploit the tested multi-threading / multi-core system, its throughput proved vastly better. Perhaps the Financial Services Front Office folk are missing a trick here?
- One fascinating project involved a Proof Of Concept (POC) that processed 500M events in 4 days, achieving up to 75K events per sec using 11 agents deployed to a 24 core server, with business control via decision tables. What surprised me was that the POC was inserted into a live operational business system – how about that for faith (or rather, reliability)!
Thanks paul. Any more info on the size of the events?
That looks nice. Since one of your older posts talks about measuring CEP performance, what would it be in those terms? I think it was a post from Nov 20th (http://www.thetibcoblog.com/2011/11/20/is-hzb-a-good-metric-for-event-processing-throughput/).
thanks
Hi Peter – good qu. I didn’t get the event size (it was telco, so presumably <1Kb and maybe as little as 10s of Bytes). So it could be anywhere from 75KHzB to 75MHzB. I’ll see about getting some more info and updates on HzB ratings for other use cases…
Cheers