Optimizing continuous queries using update propagation with varying granularities

Abstract

We investigate the possibility to use update propagation methods for optimizing the evaluation of continuous queries. Update propagation allows for the efficient determination of induced changes to derived relations resulting from an explicitly performed base table update. In order to simplify the computation process, we propose the propagation of up-dates with different degrees of granularity which corresponds to an incremental query evaluation with different levels of accuracy. We show how propagation rules for different up-date granularities can be systematically derived, combined and further optimized by using Magic Sets. This way, the costly evaluation of certain subqueries within a continuous query can be systematically circumvented allowing for cut-ting down on the number of pipelined tuples considerably

Living in Parallel Realities

“…For writing, InVerDa generates three triggers on each table version: for inserts, deletes, and updates. To not recompute all data of the materialized side after each write operation at the not-materialized side of an SMO, InVerDa adopts an update propagation technique for Datalog rules [2] that results in minimal write operations. For instance, an insert operation ∆ + T (p, A) on the table version T propagated back to the source side of a materialized SPLIT SMO results in the following update rules:…”