QScheduler: A Tool for Parallel Query Processing in Database Systems

Parallel query processing in database systems has been actively investigated for the past several years. When several queries execute at the same time, all parallel tasks will compete for access to the underlying I/O and computation resources and this will lead to more contention and thrashing. When a series of queries of different types run concurrently in a database system, if we change the execution sequence of these queries, the whole execution time of the query task may be different. This simple change in the execution order can directly affect the interactions among parallel queries. Hence, we argue that it is important to take the interactions among parallel queries into account. In this paper, we consider the problem of improving the performance of parallel query tasks based on a tool named Scheduler which is designed and implemented to control the query execution sequence. Our Scheduler treats the database system as a black box and decides which query to schedule preferentially according to the different algorithm. Our experimental evaluation using TPC-H benchmark shows that the Scheduler can capture the impact of complex query interactions and will improve the system performance efficiently.