异构集群节点与作业特性感知资源分配算法

doi:10.3778/j.issn.1002-8331.2102-0275

摘要/Abstract

摘要： 为了避免多应用间的资源争用，Spark采用了FIFO、FAIR等作业调度策略，辅以SpreadOut和非SpreadOut两种资源调度算法，但是这些算法没有充分考虑用户作业类型和集群节点性能的相互关系。用户作业类型及节点性能偏向感知的资源调度算法ATNPA提出了对该问题的解决方案。ATNPA根据作业运行所需的内存量和CPU核数将用户作业分为CPU密集型和内存密集型。节点的性能偏向性由节点的静态因素和动态因素决定。静态因素包括CPU速度、内存大小、CPU核数和磁盘容量等；动态因素包括CPU剩余率、内存剩余率、磁盘剩余率和磁盘读写速度等。ATNPA算法在进行资源分配时，能够将作业分配到最适合其类型的节点上。仿真实验表明，与未考虑节点和作业匹配的算法相比较，ATNPA能够有效缩短作业的执行时间、提高集群的性能。

关键词: 资源调度, 任务调度, 作业类型, 节点性能偏向, Spark

Abstract: To avoid resource contention among multiple applications, FIFO and FAIR are adopted by Spark, together with two resource scheduling algorithms, i.e., SpreadOut and non-SpreadOut. However, the existing algorithms do not fully consider the relationship between the user application type and the node features. To handle this issue, ATNPA（application type and node preference-aware scheduling algorithm）, an algorithm that comprehensively considers this relationship is proposed. According to its memory and CPU requirements, an application is classified into two categories, i.e., CPU intensive and memory intensive. The node preference is decided by both static attributes and dynamic attributes. The static attributes refer to the inherent features of a node, such as the memory size, the number of CPU cores, etc. The dynamic attributes are the features that may change with the working status of the node, such as the residual memory and CPU, etc. Therefore, ATNPA can allocate an application to the nodes that are most suitable to its type during the resource allocation period. Experimental results validate that compared with scheduling algorithms without considering the node preference, ATNPA can reduce the application execution time and improve system performance.

Key words: resource scheduling, application scheduling, application type, node preference type, Spark

胡亚红, 吴寅超, 朱正东, 李小轩. 异构集群节点与作业特性感知资源分配算法[J]. 计算机工程与应用, 2022, 58(18): 327-334.

HU Yahong, WU Yinchao, ZHU Zhengdong, LI Xiaoxuan. Heterogeneous Cluster Resource Allocation Algorithm Considering Application and Node Characteristics[J]. Computer Engineering and Applications, 2022, 58(18): 327-334.

参考文献

[1] Apache Hadoop[EB/OL].（2020-05-21）.http：//hadoop.apache.org/.
[2] ZAHARIA M，CHOWDHURY M，FRANKLIN M J，et al.Spark：cluster computing with working sets[C]//Proceeding of the 2nd USENIX Conference on Hot Topics in Cloud Computing，2010：1765-1773.
[3] Apache storm[EB/OL].[2021-02-21].http：//storm.apache.org/.
[4] 杨志伟，郑烇，王嵩，等.异构Spark集群下自适应任务调度策略[J].计算机工程，2016，42（1）：31-35.
YANG Z W，ZHENG Q，WANG S，et al.Adaptive task scheduling strategy for heterogeneous spark cluster[J].Computer Engineering，2016，42（1）：31-35.
[5] 廖湖声，黄珊珊，徐俊刚，等.Spark性能优化技术研究综述[J].计算机科学，2018，45（7）：7-15.
LIAO H S，HUANG S S，XU J G，et al.Survey on performance optimization technologies for spark[J].Computer Science，2018，45（7）：7-15.
[6] Apache spark[EB/OL].[2021-02-20].http：//spark.apache.org/.
[7] GUPTA S，DUTT N，GUPTA R，et al.SPARK：a high-level synthesis framework for applying parallelizing compiler transformations[C]//16th Intelligent Computing and Control Systems，2003.
[8] 徐佳俊，刘功申，苏波，等.基于Spark的异构集群调度策略研究[J].计算机科学与应用，2016，6（11）：692-704.
XU J J，LIU G S，SU B，et al.Adaptive scheduling strategy for heterogeneous spark cluster[J].Computer Science and Application，2016，6（11）：692-704.
[9] XU L，BUTT A R，LIM S H，et al.A heterogeneity-aware task scheduler for spark[C]//2018 IEEE International Conference on Cluster Computing，2018：254-256.
[10] 胡亚红，盛夏，毛家发.资源不均衡Spark环境任务调度优化算法研究[J].计算机工程与科学，2020，42（2）：14-20.
HU Y H，SHENG X，MAO J F.Research on task scheduling optimization strategy of Spark environment with unbalanced resources dissertation[J].Computer Engineering and Science，2020，42（2）：14-20.
[11] 杨忙忙.Spark数据处理平台中资源动态分配技术研究[D].北京：北京工业大学，2016.
YANG M M.Research on resource dynamic allocation technology on Spark data processing framework[D].Beijing：Beijing University of Technology，2016.
[12] 丁晶晶，张功萱.一种基于Spark在线Web服务的高效低延迟调度资源算法[J].计算机工程与科学，2016，38（8）：1550-1556.
DING J J，ZHANG G X.A high efficient and low-latency resource scheduling method for Spark on Web service[J].Computer Engineering and Science，2016，38（8）：1550-1556.
[13] WANG G，XU J，LIU R，et al.A hard real-time scheduler for Spark on YARN[C]//2018 18th IEEE/ACM International Symposium on Cluster，Cloud and Grid Computing（CCGRID），2018：645-652.
[14] CHENG D，RAO J，JIANG C，et al.Resource and deadline-aware job scheduling in dynamic Hadoop clusters[C]//Proceeding of IEEE International Parallel & Distributed Processing Symposium，2015：956-965.
[15] KUMBHARE N，MARATHE A，AKOGLU A，et al.A value-oriented job scheduling approach for power-constrained and oversubscribed HPC systems[J].IEEE Transactions on Parallel and Distributed Systems，2020，31（6）：1419-1433.
[16] ZHANG H，GENG X，MA H.Learning-driven interference-aware workload parallelization for streaming applications in heterogeneous cluster[J].IEEE Transactions on Parallel and Distributed Systems，2021，32（1）：1-15.
[17] DU H，ZHANG K，HUANG S，et al.OctopusKing：A TCT-Aware task scheduling on Spark platform[C]//2019 IEEE 25th International Conference on Parallel and Distributed Systems（ICPADS），2019：159-162.
[18] PAN F，XIONG J，SHEN Y，et al.H-scheduler：storage-aware task scheduling for heterogeneous-storage Spark clusters[C]//2018 IEEE 24th International Conference on Parallel and Distributed Systems（ICPADS），2018：9-17.
[19] MENG X，LUKASZ G.Parallel scheduling of data-intensive tasks[C]//European Conference on Parallel Processing，2020：117-133.
[20] 詹剑锋，高婉玲，王磊，等.BigDataBench：开源的大数据系统评测基准[J].计算机学报，2016，39（1）：196-211.
ZHAN J F，GAO W L，WANG L，et al.BigDataBench：an open-source big data benchmark suite[J].Chinese Journal of Computers，2016，39（1）：196-211.