Comparing global optimization and default settings of stream-based joins
Naeem, M; Dobbie, G; Weber, G
MetadataShow full metadata
One problem encountered in real-time data integration is the join of a continuous incoming data stream with a disk-based relation. In this paper we investigate a stream-based join algorithm, called mesh join (MESHJOIN), and focus on a critical component in the algorithm, called the disk-buffer. In MESHJOIN the size of disk-buffer varies with a change in total memory budget and tuning is required to get the maximum service rate within limited available memory. Until now there was little data on the position of the optimum value depending on the memory size, and no performance comparison has been carried out between the optimum and reasonable default sizes for the disk-buffer. To avoid tuning, we propose a reasonable default value for the disk-buffer size with a small and acceptable performance loss. The experimental results validate our arguments.