The StreamBase Event Processing Blog

EDA and CEP: Where's the Beef?

Jack van Hoof wrote a post this week called "EDA versus CEP, Again..." and Joe McKendrick echoed his remarks in his post "Why 'Event Driven Architecture' is more than Complex Event Processing."   Jack and Joe both suggest that CEP vendors are incorrectly trying to equate CEP with EDA.   Jack's main point is that EDA is architecture, and CEP is a tool.  My reaction was:  "Well, of course that's true!"   But what confuses me is exactly who Jack and Joe think is getting it wrong.  

I got worried for a split second, because last summer I wrote a bunch of articles about EDA, event processing, and CEP.  I wanted to make sure I hadn't been responsible for the confusion - so I went back and re-read my writing*** and I'm pretty sure Jack and I would agree on terms. Indeed, I often referred to Jack's thoughtful writing on the subject in those articles, and he commented on them.

But I think there's a point that wasn't covered by Joe and Jack - that CEP, in fact, is a really important element of an effective EDA.  It's not a required element, but it sure makes it better.  I believe that CEP will prove just as essential to EDA as a relational database is to client / server applications, or as application servers are to three-tier architecture.  And it's a fair point that because it's becoming more common in mainstream IT, that more best practices, patterns, use cases, and discourse about effective EDA is needed.

Yet, as one of the posters in Jack's blog laments:  "Why aren't more people talking about EDA, because that's what's really important?"  Perhaps EDA isn't new, and CEP is.  My first job out of school was as a developer working on an event driven trading system, in 1988.  Our group built a multicast middleware driven, real-time trading system.  The only other analagous software around was the Teknekron Information Bus, or TIB.  We used Reuter's "new" digital market data feeds, and integrated that with event-driven trading screens and applications.  It was an excellent event driven architecture.   It was new. In 1988.  We didn't use CEP because it didn't exist.  Until recently.

And perhaps that's the point:  maybe the reason why folks don't talk as much about EDA is that the disruptive element of EDA IS CEP?

*** Unfortunately Progress Apama has (I'm assuming mistakenly) attributed my writing to someboday named "Progress Apama," not the correct author: "Mark Palmer."   Regardless, the articles to which I refer (e.g., "The Myths of EDA and SOA," and "The 10 Imperatives of Algo Trading") were written by me, and I stand by them. 

CEP Begins to Break Out with Enterprise Deployment: BNY ConvergEx

A popular topic at the event processing sessions in September at the Gartner event processing conference was whether CEP would "break out" of established application areas, like algorithmic trading.   Today StreamBase announced a deployment that marks an important milestone in the industry - broad IT and business application deployment - beyond point applications.

Specifically, ConvergEx has announced it has deployed StreamBase for market interaction monitoring, routing, trading operations analysis, credit risk correlation, system health monitoring and real-time analysis of market feeds for latency and response time.  It provides these services BNY ConvergEx’s entire U.S. electronic trading infrastructure, which includes its dark liquidity pool VortEx.

CEP deployment for market interaction, routing, credit risk, and system health monitoring is a important milestone, and it shows that CEP is maturing to be a key element of the IT fabric. 

StreamBase has always been known as the high-performance, multi-threaded engine with a focus on an SQL-oriented event processing metaphor.  In my first 6 months at StreamBase, I've built an appreciation for what the StreamBase engineering team has done to make CEP truly enterprise class:  visual debugging, unit testing, profiling, and CEP coding that doesn't rely on generated code.   So I was thrilled when we announced in the release of StreamBase 6.0 - the industry's first CEP platform to deploy a visual debugger, profiler, and testing environment in June.  StreamBase became the first CEP platform to deliver on a complete set of enterprise-class deployment features.

One of the elements Joe Weisbord from ConvergEx pointed out in his quote about this deployment was the speed with which his team developed algorithms that analyze the real-time trading operations at VortEx.  This allows the ConvergEx trading operations to understand an ever-increasing set of conditions in their operations, including: “SMS me when any client order gets rejected over $50K”;  “Why was my order canceled or rejected?”;  “Which order did a reject come from?”;  “Alert me when any order on ARCA crosses my maximum tolerance for latency”;   “Warn me by email when 30 minute average latency is trending toward my maximum”;  “Alert me when any client submits an order over $150K”;  “Show me all orders rejected in the last 30 minutes and show me why?”

In the past several weeks, trading volumes have been going through the roof.  Being able to manage market interaction, routing, trading operations, credit risk correlation, and system health is becoming increasingly critical with increased volatility.  As BNY ConvergEx shows us, CEP is becoming the choice for leading trading houses to deploy mission critical controls on electronic trading.