The fix-point method for discrete events simulation using SQL and UDF

Abstract

In this work we focus on leveraging SQL's expressive power and query engine's data processing capability for large scale discrete event simulation. The challenge of using SQL query for discrete event simulation is that the simulation involves concurrent operations where the past events, the future events and the pending events (not fully instantiated yet) must be taken into account; but a SQL query is a set-wise operation that only cares of the current state. Specifically when the pending events are recursively generated as a result of processing primary events, there is no way to "insert" them back to the chronological sequence of input data, or to introduce a "clock" to the query processing. Due to these difficulties, there is no effort on using SQL query for discrete event simulation has been reported yet. We propose a mechanism for discrete event simulation, where a chronological sequence of input records, which bear pending events, are processed by a SQL query recursively until reaching the fixpoint with all the pending events instantiated. This approach is characterized by leveraging the pipelined query processing to serialize concurrent events as well as to scope event chaining, and in this way to generate concurrent events incrementally and efficiently. Our experience reveals its flexibility and scalability. © 2011 Springer-Verlag Berlin Heidelberg.