Parallel query execution in NonStop SQL

NonStop Structured Query Language (SQL) is a distributed database management system (DBMS) running on a loosely coupled multiprocessor machine. It allows horizontally partitioned relations. This environment makes it possible to execute larger operations in parallel. The different parallel execution models in NonStop SQL and the underlying process architecture are discussed, with the focus on parallel join algorithms. The current implementation and some work on future parallel execution models using hash join algorithms are described. Parallel execution in NonStop SQL is transparent to the application. Therefore, some internal component must be able to partition the work to be done for a query. One application process can utilize multiple disk processes. It would be desirable to have multivariable operations, such as joins, running in parallel. A new component, the executor server process (ESP), is introduced. The executor component in the application program (master executor) can start a set of server processes to perform queries or parts of queries in parallel. Each ESP consists of a small main program, executor, and file system. Master and server executors communicate via messages.<
>