基于超算环境的面向多租户的轻量级虚拟HPC集群的设计与实现

doi:10.11959/j.issn.1000-436x.2017279

通信学报 ›› 2017, Vol. 38 ›› Issue (Z2): 56-66.doi: 10.11959/j.issn.1000-436x.2017279

基于超算环境的面向多租户的轻量级虚拟HPC集群的设计与实现

谭郁松¹,李荣振¹,吴庆波¹,张建锋¹,张尧学^1,²

¹ 国防科技大学计算机学院，湖南长沙 410073
² 中南大学信息科学与工程学院，湖南长沙 410083

出版日期:2017-11-01 发布日期:2018-06-07
作者简介:谭郁松（1976-），男，江西余干人，博士，国防科技大学研究员，主要研究方向为操作系统、云计算与大数据。|李荣振（1985-），男，河南周口人，博士，国防科技大学博士生，主要研究方向为云计算与网络虚拟化。|吴庆波（1969-），男，浙江宁波人，博士，国防科技大学研究员，主要研究方向为操作系统与云计算。|张建锋（1984-），男，陕西宝鸡人，博士，国防科技大学助理研究员，主要研究方向为操作系统、网络安全与云计算。|张尧学（1956-），男，湖南常德人，博士，中国工程院院士，中南大学教授、博士生导师，主要研究方向为计算机网络、操作系统与透明计算。

Design and implementation of multi-tenant oriented lightweight virtual HPC cluster based on supercomputing environment

Yu-song TAN¹,Rong-zhen LI¹,Qing-bo WU¹,Jian-feng ZHANG¹,Yao-xue ZHANG^1,²

¹ College of Computer,National University of Defense Technology,Changsha 410073 China
² School of Information Science and Engineering,Central South University,Changsha 410083 China

Online:2017-11-01 Published:2018-06-07

摘要/Abstract

摘要：

为了给用户提供按需使用的 HPC 服务并解决用户应用部署的软件依赖性问题，在不破坏现有超大规模高性能集群管理的前提下，基于Fat-Tree网络拓扑和虚拟集群模型，通过对部分资源进行云化管理，设计并实现了一种轻量级的HPC集群交付模式。从而为面向租户需求的高性能应用提供一种云化的HPC租用服务，并解决了HPC应用软件栈的僵化问题，以易用的方式为更多的HPC租户提供服务。实验结果表明，该方法以近似物理节点的性能使HPC具备了云的按需使用的特性，并具有较为理想的植入效率。

关键词: 轻量级虚拟HPC, 虚拟集群模型, 胖树, 多租户

Abstract:

In order to deliver on-demand HPC services for multi-tenants and solve software dependency problems which rigidly restrict the usage of HPC applications.On the premise of not destroying the existing large scale high performance cluster management,a lightweight HPC cluster delivery model was designed and implement with the Fat-Tree network topology and the virtual HPC cluster model.So as to supply a cloud based HPC rental service for the high performance applications of the multi-tenants and solve the rigid dependable problem of HPC application software stack for more HPC tenants in an easy way,the experimental results show that the proposed method is similar to the performance of the physical nodes,so that the HPC has the on-demand characteristics of the cloud and has a better embedding efficiency.

Key words: light virtual HPC cluster, virtual cluster model, Fat-Tree, multi-tenant

中图分类号:

TP338

谭郁松,李荣振,吴庆波,张建锋,张尧学. 基于超算环境的面向多租户的轻量级虚拟HPC集群的设计与实现[J]. 通信学报, 2017, 38(Z2): 56-66.

Yu-song TAN,Rong-zhen LI,Qing-bo WU,Jian-feng ZHANG,Yao-xue ZHANG. Design and implementation of multi-tenant oriented lightweight virtual HPC cluster based on supercomputing environment[J]. Journal on Communications, 2017, 38(Z2): 56-66.

图/表 13

表1

图1

图2

图3

图4

图5

表2

图6

表3

表4

图7

图8

图9

参考文献 22

[1]	ABDELBAKY M , PARASHAR M , JORDAN K E ,et al. Enabling high-performance computing as a service[J]. IEEE Computer, 2012,45(10): 72-80.
[2]	NIU S , ZHAI J , MA X ,et al. Cost-effective cloud HPC resource provisioning by building semi-elastic virtual clusters[C]// Proceedings of SC13,Networking,Storage and Analysis. 2013.
[3]	XAVIER M , NEVES M , ROSSI ,et al. Performance evaluation of container-based virtualization for high performance computing environments[C]// PDP,21st Euromicro International Conference on, 2013: 233-240.
[4]	SLOMINSKI A , MUTHUSAMY V , KHALAF R . Building a multi-tenant cloud service from legacy code with docker containers[C]// Proc IEEE IC2E, 2015: 394-396.
[5]	MERKEL D . Docker:lightweight linux containers for consistent development and deployment[J]. Linux Journal, 2014,2014(239):2.
[6]	YOO A B , JETTE M A , GRONDONA M . Slurm:simple Linux utility for resource management[C]// Workshop on Job Scheduling Strategies for Parallel Processing. Springer,Berlin,Heidelberg, 2003: 44-60.
[7]	PANG Z , WANG K , XIE M ,et al. The TH express-2 high performance interconnect networks[J]. Frontiers of Computer Science, 2014,8(3): 357-366.
[8]	SEFRAOUI O , AISSAOUI M , ELEULDJ M . Openstack:toward an open-source solution for cloud computing[J]. International Journal of Computer Applications, 2012,55(3).
[9]	YU M , YI Y , REXFORD J ,et al. Rethinking virtual network embedding:substrate support for path splitting and migration[J]. ACM SIGCOMM Comput.Commun.Rev., 2008,38(2): 17-29.
[10]	TANTAWI A N , . A scalable algorithm for placement of virtual clusters in large data centers[C]// Proc.Int’l Symp. IEEE 20th MASCOTS, 2012: 3-10.
[11]	WEI X , LI H . Topology-aware partial virtual cluster mapping algorithm on shared distributed infrastructures[J]. IEEE Transactions on Parallel and Distributed Systems, 2013: 2721-2730.
[12]	LI X , WANG H , DING B ,et al. Resource allocation with multi-factor node ranking in data center networks[J]. Future Generation Computer Systems, 2014,32: 1-12.
[13]	GONG S Q , CHEN J , KANG Q Y ,et al. An efficient and coordinated mapping algorithm in virtualized SDN networks[J]. Front Inform Technol Electron Eng, 2016,17(7): 701-716.
[14]	PAPAGIANNI C , ANDROULIDAKIS G , PAPAVASSILIOU S . Virtual topology mapping in SDN-enabled clouds[C]// NCCA. 2014
[15]	WANG Z , HAN Y , LIN T ,et al. Virtual network embedding by exploiting topological information[C]// IEEE GLOBECOM, 2012.
[16]	LIAO J , FENG M , LI T ,et al. Topology-aware virtual network embedding using multiple characteristics[J]. KSII Trans Internet Inf Syst (TIIS), 2014,8(1): 145-164.
[17]	CHENG X , SU S , ZHANG Z ,et al. Virtual network embedding through topology-aware node ranking[C]// SIGCOMM Comput Commun. 2011: 38-47.
[18]	LI R Z , WU Q B , TAN Y S ,et al. A cost-effective approach of building multi-tenant oriented lightweight virtual HPC cluster[C]// Cloud Computing and Big Data,CCBD, 2016.
[19]	孙大为, 常桂然, 李凤云 ,等. 一种基于免疫克隆的偏好多维 QoS云资源调度优化算法[J]. 电子学报, 2011,39(8): 1824-1831.
	SUN D W , CHANG G R , LI F Y ,et al. Optimizing multi-dimensional QoS cloud resource scheduling by immune clonal with preference[J]. Acta Electronica Sinica, 2011,39(8): 1824-1831.
[20]	张尧学, 周悦芝 . 一种云计算操作系统TransOS:基于透明计算的设计与实现[J]. 电子学报, 2011,39(5): 985-990.
	ZHANG Y X , ZHANG Y X , ZHOU Y Z . A new cloud operating system:design and implementation based on transparent computing[J]. Acta Electronica Sinica, 2011,39(8): 1824-1831.
[21]	程祥, 张忠宝, 苏森 ,等. 虚拟网络映射问题研究综述[J]. 通信学报, 2011,32(10): 143-151.
	CHENG X , ZHANG Z B , SU S ,et al. Survey of virtual network embedding problem[J]. Journal on Communications, 2011,32(10): 143-151.
[22]	江逸茗, 兰巨龙, 程东年 ,等. 分布式环境中基于协商的虚拟网映射算法[J]. 通信学报, 2014,35(12): 62-69.
	JIAN Y M , LAN J L , CHENG D N ,et al. Virtual network embedding algorithm based on negotiation in distributed enlivenment[J]. Journal on Communications, 2014,35(12): 62-69.

符号	解释
G_s	物理集群
N_s	物理集群节点
L_s	物理集群链路
A_n	物理集群节点属性
A_l	物理集群链路属性
P_s	物理集群路径
G_v	虚拟集群
V_v	虚拟集群节点
E_v	虚拟集群链路
R_v	虚拟机器节点资源需求
R_e	虚拟机器链路资源需求
f_n	虚拟节点到物理节点的映射函数
f_l	虚拟链路到物理路径的映射函数

硬件设备	天河二号计算节点
CPU	Intel Xeon E5-2692 2.20 GHz X 2
Memory	64 GB
存储盘阵	12.4 PB
网络设备	THNI42 Gps

OS	Ubuntukylin x64
Docker Engine	Docker version1.90
KylinCloud	V4.0（openstack Juno）
MPI	xxx
Luster	xxx
Slurm	xxx

平台/测试工具	Linpack/s	网络带宽(MB/s)
裸机	359.70	6 602.34
虚拟HPC集群	371.66	6 524.16
性能损耗	3.32%	1.18%

基于超算环境的面向多租户的轻量级虚拟HPC集群的设计与实现

Design and implementation of multi-tenant oriented lightweight virtual HPC cluster based on supercomputing environment

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 13

参考文献 22

相关文章 8

Metrics

推荐阅读 0

[1]	曹利峰,卢新,高振升,杜学绘. 基于L-DHT的多租户虚拟域隔离构建方法[J]. 通信学报, 2020, 41(6): 184-201.
[2]	王淼,王利明,徐震,马多贺. 基于熵变的多租户云内DDoS检测方法研究[J]. 通信学报, 2016, 37(Z1): 204-210.
[3]	李晓娜，李庆忠，孔兰菊，闫中敏. SaaS多租户数据放置的优化调整策略研究[J]. 通信学报, 2014, 35(Z2): 10-71.
[4]	李晓娜,李庆志,孔兰菊,闫中敏. SaaS多租户数据放置的优化调整策略研究[J]. 通信学报, 2014, 35(Z2): 63-71.
[5]	李晓娜,李庆忠,孔兰菊,庞成. 基于共享模式的SaaS多租户数据划分机制研究[J]. 通信学报, 2012, 33(Z1): 110-120.
[6]	任怡,张菁,陈红,吴庆波,孔金珠,戴华东,管刚. 云应用引擎的资源监控和计费机制研究[J]. 通信学报, 2012, 33(Z1): 192-200.
[7]	马坤,杨波,陈贞翔,马炳先,李庆忠. 从传统的Web应用到多租户SaaS应用的转换方法[J]. 通信学报, 2011, 32(9A): 133-138.
[8]	张秀伟,何克清,王健,李征. SaaS服务超市的构建模型和服务推荐方法[J]. 通信学报, 2011, 32(9A): 158-165.