面向分布式计算环境的接纳控制机制研究

接纳控制算法是分布式计算环境中资源提前预留机制的核心。本文研究总结了现有的灵活预留接纳控制算法,并在此基础上提出一种新的算法——可迁移提前预留接纳控制算法。对于不可拓展的固定请求,它允许通过迁移已预留的可拓展请求,为固定请求空出足够的资源,从而提高固定请求被接纳的概率。性能优良的数据结构对于资源预留有着至关重要的作用,合理的数据结构能显著改善接纳控制的总体性能,本文总结了现有的数据结构,包括时隙数组、时隙线段树、资源树、二叉查找树等。通过与三种经典的可拓展预留接纳控制算法的对比实验表明,可迁移提前预留能改善固定请求接纳率及总请求接纳率。

Research on Admission Control Mechanism for Distributed Computing Environments

Admission control algorithm is the core of resource advance reservation mechanisms for distributed computing environments.This paper summarizes the existing flexible reservation admission control algorithms and proposes a novel algorithm-admission control algorithm with transferable reservation in advance.The algorithm can transfer the scalable requests that have been reserved to reserve enough resources for fixed requests,increasing the acceptance probability of fixed requests.Data structure that stores the reservation information acts as an important role in resource reservation.Reasonable data structure can significantly improve the overall performance of admission control.Therefore,this paper further summarizes the existing data structure,including the slot array,the slot segment tree,the tree of resources and the binary search tree.The comparison experiments with three kinds of transferable reservation admission control algorithm show that the transferable reservation in advance algorithm can improve the acceptance rate of fixed requests and the total request accepted rate.