Automatic Parallelization of Sequential Queries

We suggest the use of transaction concurrency control paradigms to perform dynamic dependency analysis of a database query written in sequential code, in order to execute it in parallel. The sequential code is divided into small portions, and these portions are concurrently executed as separate "transactions". By using concurrency control mechanisms, we ensure that the concurrent execution of the portions is logically equivalent to the sequential execution of the original query. We present a low-overhead order-preserving concurrency control strategy to execute concurrently the nested transactions that are generated by the parallelized execution. We present performance figures to from an implementation to indicate the potential benches of our strategy.