28 August 2008 News

 

StreamBase and Oracle aim for CEP language standard

Complex event processing pioneer StreamBase has teamed with researchers from Oracle Corporation to research CEP techniques and work on standard language implementation issues in an effort to support the growth of CEP by applying standards that will allow enterprise developers to quickly understand, learn, and apply stream processing technologies.

The pair will be presenting a paper, Towards A Streaming SQL Standard, that is the result of the nearly two years of research, at the Very Large Data Base Conference in Auckland, New Zealand on 28 August.

At the conference, representatives from StreamBase and Oracle will discuss creating one common standard. One of the central observations in the paper is its characterisation of event-based, versus set-based event processing models.  StreamBase's execution model is event-based, which is useful for deterministic event-by-event data processing, where the outcome of business logic is determined by the temporal order of the data. Oracle's time-based model processes data in sets - an approach optimised for processing stored event data. A standards-based approach that embraces both models could result in an event processing approach that addresses both real-time and historical event processing in one language.

"Converging on language standards is good for the entire event processing infrastructure, from end-users to vendors to the large ecosystem of streaming data providers and application partners.  Standardisation makes programming more portable and allows greater development of third-party applications," said StreamBase co-founder and chief architect, Richard Tibbetts, who is co-presenting the paper at the conference. "Our goal is to focus on the expanding business problems CEP solves, and compete on the merits of the CEP platform, rather than differences in programming language."

Oracle and StreamBase researchers proposed that the best solution for event-based and set-based language approaches to converge would be to cleanly integrate the two standards with unifying extensions.  For example, a proposed language operator called SPREAD would allow data sets to be aggregated with ordering imposed.  Such a technique would allow a natural fusion of past and real-time event data.  The details and semantics of which are described in the joint research team's conference paper, which will be available from 28 August at www.streambase.com.