ATLAS: a Small but Complete Extension of SQL for Data Streams and Data
Mining

We propose simple extensions of SQL to support (i) continuous queries
on streaming data, and (ii) complex applications on  persistent data
stored in a database.  The linchpin of our approach is a native
extensibility mechanism for SQL based on user-defined aggregates
(UDAs) and table functions. This makes ATLAS Turing complete---without
coding external functions or embedding queries in a procedural
language.  Because of this power, and the stream-oriented computation
model of UDAs, very complex continuous queries on streams can be
supported with minimal extensions to SQL.  Furthermore, the notion of
safe and non-blocking computations on infinite data streams can be
formally characterized in this context.  The ATLAS prototype
optimizes response time and/or throughput on data streams taking
users' input into account.