Abstract:

The distributed stream architecture extends the stream model in the distributed environment, which can provide an efficient and lowcost execution environment for big data computing applications in the Internet. Meanwhile, the long communication overhead in the Internet restricts the performance of the distributed stream architecture. We propose a data stream Eager transmission technique and implement it in a prototype system of the distributed stream architecture, which improves the performance of applications by exploring the parallelism between communications and computation, as well as hides communication latency. Experimental results show that the proposed technique averagely gains a 19.58% reduction of the execution time cost, significantly improves the performance of the applications and has a promising prospec.

Key words: distributed stream architecture;performance optimization;stream Eager transmission technique