Geostreaming at IBM: infosphere streams and intelligent traffic systems

Based on a system developed at IBM Research over eight years, and in production in one of the most demanding environments imaginable for the last three, IBM has introduced InfoSphere Streams, an advanced, commercial stream processing platform. With an innovative, distributed-processing runtime and a graph-based, extensible programming paradigm, it is well suited to extreme performance requirements and highly sophisticated processing and analytics on all kinds of data, from structured business records to text, audio, imagery, time-series and geospatial data. It is already being used in numerous scientific applications and academic projects. In this talk, I will give an overview of the Streams product and then highlight one such project, a collaboration with the KTH University in Stockholm. The ultimate aim of the project is to build an intelligent route planning and travel time prediction system based on instantaneous information about speeds and travel times from GPS-instrumented vehicles as well as historical data collected by the same vehicles. I will discuss the design and implementation of the application as well as performance results from initial simulations.