Monday, November 27, 2006

When is ESP not ESP?

Whenever a new concept in technology makes it onto the radar screens of analysts and a few forward thinking companies, vendors tend to want to share in some of the limelight. That is why, for example, after Gartner analysts coined the phrase, Enterprise Service Bus (ESB) and it gained significant notoriety and publicity, even mainstream EAI vendors that initially rejected the ESB concept, were champing at the bit to say, 'We've got an ESB offering too.'

It appears that the same thing may be happening in the event stream processing (ESP) space. In my last post I covered off on ESP, a relatively nascent market, and how it was being used in trading applications, logistics and company supply chains to enable companies to respond and act on real-time streaming data and events.

A word of warning, however, is that as ESP is a relatively immature market, definitions of what constitutes ESP differ from vendor to vendor. Phil Howard, research director at Bloor Group, defines an event, as "an event of some importance." In other words, an event stream processing application is not interested in every event that may occur.

Events for example, can come from transaction databases, Bloomberg or Reuters market data feeds or RFID tags on boxes of books. Event processing is also about managing exceptions such as credit card fraud detection. The next step up from that is complex event processing (CEP), which Howard says is managing 'a set of different exceptions.' "It is easiest to think of ESP as a pipe with water flowing through it and onto that pipe are placed fine mesh grills," Howard explains. "The water flows through those mesh grills, which are not fixed but interchangeable."

In its SOA maturity model, Oracle apparently puts CEP at Level 5, indicating that for most companies it is something that they consider implementing much later on if at all.

However, as Howard points out, firms can implement CEP without having to go down the service-oriented architecture route. In algorithmic trading, for example, which uses event stream processing to detect movements in stocks based on pre-configured algorithms, firms have not necessarily implemented an SOA.

Whilst event stream processing is about handling throughput of data, when it comes to complex event processing, Howard says it is all about implementing technology that can handle complex data streams. Traditional relational databases are less favored in this environment as the perception is that they fall far short of the requirements for responding to incoming data streams in a timely fashion.

By now you are probably thinking isn't ESP or CEP just another form of business intelligence? Well, yes of sorts. According to Howard, event processing incorporates real-time operational business intelligence. However, he adds, some of the core business intelligence software vendors such as Business Objects, have technology that is not "process aware", which is needed if companies want to build operational business intelligence platforms based on CEP or ESP.

Howard says some of the database vendors are looking to embed more intelligence into their data warehousing offerings. He cites the example of Sybase, which he says is looking to partner with an event processing vendor on the front end so it can offer a complete solution. IBM's WebSphere Front Office for Financial Markets allows companies to combine and filter data feeds, and although it may act as a front-end to an event processing engine, according to Bloor Group it is not an event processing solution as such. Click here for more of Bloor's insights.

No comments: