Abstrakt
Since the development of applications for parallel architectures is complicated and error-prone, many frameworks were created to simplify this task. One promising approach which is applicable especially for the development of parallel databases is expressing algorithms as stream programs, i.e. inputs and outputs of procedures are data streams and these procedures are connected so that they form an oriented graph.
In this paper, we introduce highly scalable sorting algorithm which is suitable for streaming systems. We achieve this mainly by introducing multiway merge algorithm which is able to merge multiple independent sorted streams in parallel.