Grouped Processing of Relational Algebra Expressions over Data Streams

摘要

Implementation of the data stream processing applications requires a method for formal specification of the computations at a dataflow level. The logical models of stream processing hide the lower level implementation details. To solve this problem, we propose a new model of data stream processing based on the concepts of relational data stream, extensible system of elementary operations on relational streams, and data stream processing network integrating the dataflows and elementary operations. Next, we present the transformations of grouped data stream processing applications into data stream processing networks. The transformations proposed in the paper integrate the networks and optimize the implementations through elimination of the redundant elementary operations and dataflows. Finally, the paper introduces a timestamp based synchronization of data flows in our model and discusses its correctness.