摘要近来,云计算技术在工业界和学术界共同推动下取得了巨大的发展,大量的云计算框架投入使用,为各个框架分配资源的集群资源管理器也应运而生。而资源调度技术作为集群资源管理器的核心技术之一,其主要功能是对框架的执行顺序和计算资源的分配进行控制,这直接关系到集群的整体性能和资源的利用情况。本文在深入了解Mesos平台下调度器的运行机理的基础上,实现先进先出、轮询、大作业优先这三种经典的调度策略,详细分析了三种调度策略在Mesos上的具体实现。通过对这三种策略的正确性实验和性能对比实验,分析其实际效果与理论效果的差距。最后得出这三种策略在集群环境下的优缺点。80431
毕业论文关键词 Mesos 资源调度 先进先出 轮询 大作业优先
毕业设计说明书外文摘要
Title Resource Scheduling Algorithm’s Research for Distributed Systems
Abstract Recently ,cloud computing has achieving tremendous development promoted by industry field and academy ,more and more cloud computing Frameworks have been put into service ,and the cluster resource management to allocating resources for each Framework also emerged 。Resource scheduling technology is one of cluster resource management’s key technologies ,its main function is to control a Framework’s execute sequence and computing resources’ distribution ,which directly relates to cluster’s overall performance and cluster resources’ usage 。based on understanding its operational mechanism of the scheduler on the Mesos platform deeply ,this thesis implements three classic scheduling polices which are FIFO,POLL and big job priority scheduling policy 。Firstly, analyzes the implementation of three scheduling policies on Mesos amply 。Secondly, analyzes gaps in their practical effect with the theoretical effect base on results of the correctness of these three polices and performance comparison experiments。 Finally ,Concludes these three scheduling polices’ pros and cons in the clustered environment。
Keywords Mesos Resources Scheduling FIFO POLL Big Job Priority
目 次
1 绪论 1
1。1 问题背景与研究意义 1
1。3 论文研究内容 1
2 Mesos相关概念与运行机制 3
2。1 Mesos总体架构介绍 3
2。1。1 Mesos-Master介绍 3
2。1。2 Mesos-Slave介绍 3
2。1。3 Framework介绍 3
2。2 Mesos组件间的联系与通信 4
2。3 Mesos资源调度流程 4
3 Mesos调度器的层次分析与设计 6
3。1 Mesos默认调度器的实现层次 6
3。2 设计自定义调度器的实现层次 7
3。3 自定义调度器需实现的内容 7
4 自定义调度器的设计与实现 9
4。1 先进先出算法的设计与实现分析 9
4。1。1 FIFO算法流程图