Progressive processing for querying system behavior
Methods for querying a database and database systems include optimizing a database query for parallel execution using spatial and temporal information relating to elements in the database, the optimized database query being split into sub-queries with sub-queries being divided spatially according to host and temporally according to time window. The sub-queries are executed in parallel. The results of the database query are outputted progressively.
1. A method for querying a database, comprising:
optimizing a database query that includes an update frequency for parallel execution using spatial and temporal information relating to elements in the database, by splitting the database query into sub-queries in accordance with a sequential workload partitioning with initialization cost strategy, computing initialization costs as separate workloads for the purpose of partitioning, the optimized database query being split into a plurality of sub-queries that have an expected execution time based on the update frequency, with sub-queries being divided spatially according to host and temporally according to time window;
executing the sub-queries in parallel;
determining actual execution information for executed sub-queries;
adjusting an event processing rate for subsequent sub-queries based on the determined actual execution information for the executed sub-queries; and
outputting progressive results of the database query according to the update frequency.
2. The method of claim 1 , wherein the database query further comprises a subject, an operation, and an object that the subject operates on.
3. The method of claim 1 , wherein outputting progressive results of the database query comprises outputting results from executed sub-queries.
4. The method of claim 1 , wherein the sequential workload partitioning with initialization cost strategy is online adaptive workload prediction partitioning.
5. A database system, comprising:
a query optimizing module comprising a processor configured to optimize a database query that includes an update frequency for parallel execution using spatial and temporal information relating to elements in the database, to split the database query into sub-queries in accordance with a sequential workload partitioning with initialization cost strategy, to compute initialization costs as separate workloads for the purpose of partitioning, the optimized database query being split into a plurality of sub-queries that have an expected execution time based on the update frequency, with sub-queries being divided spatially according to host and temporally according to time window, to determine actual execution information for executed sub-queries, and to adjust an event processing rate for subsequent sub-queries based on the determined actual execution information for the executed sub-queries;
a parallel execution module configured to execute the sub-queries and new sub-queries in parallel; and
a results module configured to output progressive results of the database query according to the update frequency.
6. The system of claim 5 , wherein the database query further comprises a subject, an operation, and an object that the subject operates on.
7. The system of claim 5 , wherein the results module is further configured to output progressive results from executed sub-queries.
8. The system of claim 5 , wherein the sequential workload partitioning with initialization cost strategy is online adaptive workload prediction partitioning.