Ben Whittle, UBS

Imagine a world of continuous automated pattern recognition. Hundreds of thousands of events must be observed and processed every second of every day, just to keep up.This is the world of the modern trading floor. This talk will look at the architectural implications of real time event processing. In particular we shall consider how the introduction of the dimension of time changes how we see data, queries, and relationships:

• Data changes more frequently, affecting distribution and isolation
• Queries are on event streams, allowing more general pattern matching
• Relationships express complex causality and behaviour, and not simple I/O constraints

We shall then show how this affects use cases in practice. We look at when a database is a good solution for data, queries, and relationships in the new world, and when it is not. We shall also provide examples of cases where performance and behaviour may not be as predicted, and consider emerging tools and patterns for addressing this problem.