基于数据交换与同步的作业调度方案

多机作业调度问题是一个经典的NP难问题,在应用中由于实际需要,会出现各种约束和变形,调度问题的研究成果决定着系统的性能.DataTurbo是作者参与的一个用于解决分布式数据迁移、集成和融合的平台,该平台承担着大数据量的分布式传输任务.在DataTurbo平台基础上,提出一种适用于数据交换与同步的分布式作业调度方案,并构建一个灵活的分布式调度算法框架,解决相关的调度问题.该调度方案是一种在线的、可并发的、作业可分解的多机调度方案.仿真实验结果显示,该调度方案在任务负载大、调度点稀疏情况下优势明显,能适用于数据交换同步作业,可作为数据交换与同步作业的动态调度方案,并为相关启发式算法建立基础模型.

Job scheduling solution based on data exchange and synchronization

Multiprocessor scheduling is a classical NP-hard problem. However, in actual cases, the scheduling research, with variable constraints and transformations, is so significant and determines system performance. DataTurbo is a platform used to migrate, integrate and syncretize distributed data. The authors proposed a distributed task scheduling solution adaptive to data exchange and synchronization based on DataTurbo, which took massive data transferring tasks. Also, the authors constructed a flexible distributed scheduling algorithm framework to solve related scheduling issues. The solution is an online, concurrent multiprocessor scheduling solution, whose tasks can be separated. The experimental results show that the solution works well with obvious advantage in jobs which have tasks with large load and sparse scheduling points, so that it can provide dynamic scheduling solution for data exchange and synchronization and establish foundation model for relevant heuristic algorithm.