云—边缘系统中跨域大数据作业调度技术研究.

In order to minimize the average completion time of a series of geo-distributed data analytical jobs, this paper first introduced online random scheduling algorithm ranT A. ranT A actually showed the preference on the task assignment under the consideration of both computing capacity of edges and the network bandwidth. Furthermore, in order to avoid overloading those edges with low computing capacities, this paper proposed data redistribution mechanism ranT A-data based on ranT A by redistributing some data to the central data center along with the tasks. The result of ranT A -data could be proved concentrated on its optimum with high probability. Extensive simulations show that ranTA-data improves nearly 30% performance compared with traditional algorithms. [ABSTRACT FROM AUTHOR]