数据操作系统中基于博弈的多资源分配算法A Multiple Resource Allocation Algorithm Based on Game Theory in DataOS
张潇,王凯,武延军
摘要(Abstract)：
数据操作系统需要对CPU和内存等多种系统资源进行管理,为了在公平的前提下,解决不同用户对上述资源的不同需求问题,本文基于完全信息的动态博弈提出了ICEEI算法。该方法通过构建博弈树并优化博弈树的方法解决数据操作系统中的多资源分配问题。该算法最符合实际的假设是认为任务是不可分的,即数据操作系统分配给每个用户的资源可完全满足任务的需求。对该算法的公平性进行了讨论,指出其满足共享激励和Pareto有效等性质。通过一系列的仿真实验,证明ICEEI可以很好地应对用户对资源需求的动态变化,并且与DRF相比在有些情况下具有更高的资源利用率。
关键词(KeyWords)： 博弈论;多资源分配;数据操作系统
基金项目(Foundation): 中国科学院战略性技术先导专项“面向感知中国的新一代信息技术先导专项(XDA06000000)”资助
作者(Author): 张潇,王凯,武延军
参考文献(References)：
- [1]A.Ghodsi,M.Zaharia,B.Hindman.Dominant resource fairness:fair allocation of multiple resource types.USENIX NSDI.2011.
- [2]M.Schwarzkopf,A.Konwinski,M.Abd-El-Malek.Omega:flexible,scalable schedulers for large compute clusters.Proceedings of the 8th ACM European Conference on Computer Systems.ACM,2013:351-364
- [3]H.R.Varian.Equity,envy,and efficiency[J].Journal of economic theory,1974,9(1):63-91
- [4]J.F.Nash Jr.The bargaining problem[J].Econometrica:Journal of the Econometric Society,pages 1950,155-162
- [5]D.Fudenberg and J.Tirole.Game theory[M].MIT Press,1991.
- [6]Parkes D C,Procaccia A D,Shah N.Beyond dominant resource fairness:extensions,limitations,and indivisibilities.Proceedings of the 13th ACM Conference on Electronic Commerce.ACM,2012:808-825
- [7]W.Wang,B.Li,and B.Liang.Dominant resource fairness in cloud computing systems with heterogeneous servers[J].arXiv preprint arX-iv:1308.0083,2013.
- [8]M.Isard,V.Prabhakaran,J.Currey等.Quincy:fair scheduling for distributed computing clusters.Proceedings of the ACM SIGOPS22nd symposium on Operating systems principles.ACM,2009:261-276
- [9]Cluster scheduler simulator,https://code.google.com/p/cluster-scheduler-simulator/.
- [10]B.Hindman,A.Konwinski,M.Zaharia.Mesos:A platform for fine-grained resource sharing in the data center.Proceedings of the8th USENIX conference on Networked systems design and implementation.USENIX Association,2011:22-22