Migration-Based Online CPSCN Big Data Analysis in Data Centers

It is critical to schedule online data-intensive jobs effectively for various applications, including cyber-physical-system and social network system. It is also useful to support timely decision making and better prediction. In this paper, we investigate the online job scheduling problem with data migration for global job execution time reduction. We first establish a time model based on the real experimental results, and propose an online job placement algorithm by taking into account the benefit of both instantaneity and locality for the jobs. We then introduce data migration to the job placement algorithm. The core idea is to make a tradeoff between the migration cost and remote access cost. The simulation results demonstrate that our algorithm has a significant improvement than FIFO, and data migration shows effectiveness on global job execution time reduction. Our algorithms also provide an acceptable fairness for jobs.