Abstract
Many systems for big data processing have been developed to analyze and process huge amount of data. One of them is Storm, whose target is stream data processing. The default scheduler in Storm uses the round robin method to assign tasks, which is not optimal for heterogeneous computing environments. In this paper, we proposed and implemented a new Storm scheduling algorithm, named G-Storm, which takes the GPU capacity into consideration and can make better use of GPU to speed up the overall performance. The experimental results show that G-Storm can achieve 1.65x to 2.04x performance improvement on lightly weight and heavily loading of topology, comparing to the results of the original Storm scheduler.