IBM, BEP and CEP

IBM has just held its first annual analyst’s conference on what it calls Business Event Processing (BEP). Now, followers of the complex event processing (CEP) market will know that IBM has renamed its acquisition of AptSoft (a CEP product) as WebSphere Business Events so you could be forgiven for thinking that IBM was simply redefining CEP as BEP. However, that is by no means the case: the company defines business event processing much more widely than just CEP.

To give you an idea of where IBM’s thinking is, there were three sets of presentations: on how you use WebSphere Business Events in conjunction with business process management; on how Tivoli uses event monitoring across both IT and other asset environments; and on the introduction of IBM’s own (new) CEP engine, InfoSphere Streams. This actually made the conference sort of confusing because these areas tend to be covered by different analysts, each of whom was only interested in a part of the proceedings. However, leaving that aside, the point is that IBM sees BEP much more broadly than CEP and, indeed, it claims to be the market leader in BEP with more than 3,700 customers. I have no way of disputing, or even checking, this.

As regular readers will appreciate my main interest here was on InfoSphere Streams, which the company is actually positioning as a “stream computing engine used for CEP deployments”. Quite why the company is making this distinction is not clear to me (though I will speculate about it in a separate article). In any case, my information is limited at present but I can give you a flavour. The first thing is that the product has been designed for very high throughput, low latency, highly complex environments. In particular, it is hardware agnostic which means that not only will it run on your average server but it will also run on IBM’s supercomputers. This means that you can really get extreme performance out of it: for example, IBM claims that it is running at 2 million transactions per second but that one of its early adopters expects to have it running at 5 million transactions per second with a sub-millisecond latency. That is seriously impressive, Streambase and Progress: eat your hearts out!

InfoSphere Streams comes with a development language (and compiler) called SPADE (stream processing application declarative engine) that has been specifically developed for processing streams. However, it could equally well be described as an ADE (application development environment—and it was, at the conference) as it includes tools for things like debugging as well as the language itself.

This raises an interesting issue with respect to other IBM approaches to CEP. For example, WebSphere Business Events uses a more 4GL-like approach with no coding, while the company is also involved in the formulisation of StreamSQL as a standard, along with Streambase, Oracle, Coral8 and Truviso. Quite how all of these might pan out remains to be seen but clearly IBM wants to keep its options open in the event of any agreed standard appearing.

Anyway, back to SPADE. While IBM has plans to add graphical application composition features that could be used by business analysts in the future, at present SPADE is purely for developers. This is unusual as most CEP vendors offer both. Initial comments from early adopters suggest that it is an easy environment to work in so this may not be too much of an issue but I haven’t seen it yet so I can’t comment. One possibility that IBM has been exploring is to front-end InfoSphere Streams with WebSphere Business Events (which is more focused at the business level), using the former to determine the events that are actionable or to do deep analysis and the latter to determine the relevant processing by applying relevant business rules.

This raises another point, which is the integration across IBM’s portfolio in support of events. This is across a broad front and, in keeping with its general view of the importance of events it is event-enabling various existing products. So, for example, it is going to release a (free) support pack for CICS that will generate events from CICS transactions that can then be processed using one or other of its event processing technologies.

IBM has announced various new products and extensions to products and there will be more to come. IBM clearly sees event-driven architectures becoming pervasive and it wants to be able to play in all areas of this market, with systems running on laptops up to those on supercomputers. While its entrance into this sector will be heralded by the likes of Streambase and Progress as validating the market, and should give them some temporary momentum, it is quite clear that IBM is aiming to dominate this market: it has put a stake in the ground and it is a big (blue) stake.