A preemptive truthful VMs allocation online mechanism in private cloud
Introduction
Cloud computing has been employed by HPC and technical computing data centers to maximize return on investment (ROI). HPC system managers can leverage the benefits of a cloud model for their traditional HPC environments in order to improve scalability, simplify service access, accelerate collaboration or funding, enable pay-for-use, and improve efficiency. Many HPC clouds assume the form of private Infrastructure as a Service (IaaS). In private clouds, fixed price is easy to understand and primarily employed, but it is not economically efficient. By contrast, auction-based mechanisms are feasible solutions for virtual machines (VMs) allocation in clouds. An example of an auction-based VMs allocation mechanism is spot instances in Amazon EC2 [1]. This study focuses on auction-based mechanisms for cloud VMs allocation.
Most research has focused on resource allocation in offline situations, which do not consider the users’ random arrival and departure at any time over a long period. Users often arrive randomly and must leave at a particular moment, which can be described as the users’ lifetime. If sufficient cloud resources are allocated to the users in their lifetime, they can obtain a value that expresses payoff when a task is completed. This value is a user's private information. Our online mechanism is an appropriate response to the aforementioned situations. Designing a VMs allocation mechanism in a private cloud that maximizes the total value of the users is our primary challenge because of self-interested users and their random arrival.
In this study, we present a truthful preemptive VMs allocation online mechanism, and compare our mechanism with the optimal offline mechanism through experiments. Results reveal a good actual competitive ratio. In addition, we analyze the relationship between the performance and resource capacity. Our main contributions are the following: our online mechanism addresses the actual situation of a private cloud; truthfulness, which is the most important property, is satisfied; and the social welfare is high by the real data simulations.
This study is organized as follows. Related studies are described in Section 2. Section 3 describes our modeling of the preemptive VMs allocation online mechanism in cloud. Section 4 describes our online mechanism, for which we prove its truthfulness. In addition, allocation and payment algorithms are described. Section 5 describes the numerical experiments conducted to evaluate the performance of our online mechanism. We conclude this study in Section 6.
Section snippets
Related work
A mechanism comprises a group of rules used for an aggregative outcome in which the participants are self-interested with the private information about their preferences. Mechanism design aims to identify the good system-wide means to obtain true participant preferences regarding outcomes. A mechanism is mainly used in the field of microeconomics and resource allocation. It is also used in distributed artificial intelligence and communication networks. In [2], a VCG-Kelly mechanism for
VMs allocation online mechanism model
Providers of cloud computing provision VMs to users and aspire to maximizing revenue, utilizing resources, social welfare, and/or other objectives. Social welfare refers to the sum of value obtained by each user under a certain allocation. Private clouds are constructed by certain enterprises, institutions, and organizations. The users are limited in the internal members. Private clouds have the advantages such as rapid deployment and resources customization because the same principle applies
Mechanism description
A mechanism includes two rules: allocation and payment. Based on the VMs allocation online mechanism model, our mechanism M is given by the following rules:
Allocation rule: compute every effective user's priority valuewhere is the implemented length of user i's task up to t and λ is a constant between 0 to 1 in a certain mechanism. All are sorted in descending order, the first C users are each allocated one VM respectively at unit t. If several are equal, user i
Performance evaluation
To the best of our knowledge, publicly available cloud computing workloads are not currently available. Our simulation data were obtained from [28]. These included many workload logs from large scale parallel systems from various countries throughout the world. We used the RICC-2010 log, which derives from the RIKEN integrated cluster of clusters (RICC) that has operated since August, 2009. RIKEN is an independent scientific research and technology institution of the Japanese government.
More
Conclusions
HPC on cloud often employs pay-for-use as a resource strategy. Thus, an auction-based VMs allocation online mechanism is a promising solution for HPC cloud resource allocation-related problems, which are based on the randomly arriving and self-interested users. Promoting social welfare in private cloud is a reasonable goal, but user private information must be revealed. Therefore, in this study we proposed a novel truthful preemptive online mechanism for VMs allocation. We compared our
Acknowledgments
The workload log from the RICC cluster was graciously provided by Motoyoshi Kurokawa. This work is supported by the National Natural Science Foundation of China under Grant Nos. 61170029, 61472240, and 61373032, and the Zhejiang Provincial Science and Technology Plan of China under Grant No. 2013C31097.
Dr. Yonggen Gu received the Ph.D. and Master degree in computer science from Shanghai Jiao Tong University in 2006 and 1996. He received the Bachelor degree in mathematics from Hangzhou University in 1990. He is currently a professor in School of Information Engineering, Huzhou University. His current research interests include game theory, mechanism design and cloud computing.
References (28)
- http://aws.amazon.com/ec2/spot-instances, (accessed September...
- et al.
VCG-kelly mechanisms for allocation of divisible goods: adapting VCG mechanisms to one-dimensional signals
IEEE J. Sel. Areas Commun.
(2007) - et al.
An efficient Nash-implementation mechanism for network resource allocation
Automatica
(2010) - et al.
Hierarchical auction mechanisms for network resource allocation
IEEE J. Sel. Areas Commun.
(2012) - et al.
Double auction mechanisms for resource allocation in autonomous networks
IEEE J. Sel. Areas Commun.
(2010) - et al.
G-Hadoop: MapReduce across distributed data centers for data-intensive computing
Future Gener. Comput. Syst.
(2013) - et al.
Energy-aware parallel task scheduling in a cluster
Future Gener. Comput. Syst.
(2013) Injection based dynamic power management and an optimal policy for multiprocessor systems
Int. J. Network. Comput.
(2012)- et al.
Rollup: non-disruptive rolling upgrade with fast consensus-based dynamic reconfigurations
IEEE Trans. Parallel Distrib. Syst.
(2015) - et al.
Dynamic task flow scheduling for heterogeneous distributed computing: algorithm and strategy
IEICE Trans. Inf. Syst.
(2007)
A case for cooperative and incentive-based coupling of distributed clusters
Heuristics and evaluations of energy-aware task mapping on heterogeneous multiprocessors
An efficient green control algorithm in cloud computing for cost optimization
IEEE Trans. Cloud Comput.
An optimal control policy to realize green cloud systems with SLA-awareness
J. Supercomput.
Cited by (5)
Performance evaluation of communications in distributed systems and web based service architectures
2017, Journal of Computational ScienceControl frameworks in network emulation testbeds: A survey
2017, Journal of Computational ScienceCitation Excerpt :The purpose of resource allocation mechanism is to solve optimization problems of infrastructure utilization. For examples, Gu et al. [51] proposed a preemptive virtual machine allocation mechanism for system optimization. Balis et al. [52] also shared their views on how to use the partition tree for satisfying the computing performance in cloud resource allocation.
Recent advances in parallel techniques for scientific computing
2016, Journal of Computational ScienceVCG auction based idle instance bidding to increase IaaS provider’s profit in hybrid clouds
2017, Communications in Computer and Information Science
Dr. Yonggen Gu received the Ph.D. and Master degree in computer science from Shanghai Jiao Tong University in 2006 and 1996. He received the Bachelor degree in mathematics from Hangzhou University in 1990. He is currently a professor in School of Information Engineering, Huzhou University. His current research interests include game theory, mechanism design and cloud computing.
Jie Tao received the Master degree in information and communication engineering from Fudan University in 2007. He received the Bachelor degree in Electronic Engineering from Hangzhou University in 1997. He is currently an associate professor in School of Information Engineering, Huzhou University. His current research interests include game theory, mechanism design and cloud computing.
Dr. Guoqiang Li received the B.S., M.S., and Ph.D. degrees from Taiyuan University of Technology, Shanghai Jiao Tong University, and Japan Advanced Institute of Science and Technology in 2001, 2005, and 2008, respectively. He worked as a postdoctoral research fellow in the graduate school of information science, Nagoya University, Japan, during 2008–2009, as an assistant professor in the school of software, Shanghai Jiao Tong University, during 2009–2013, and as an academic visitor in the department of computer science, University of Oxford during 2015–2016. He is now an associate professor in school of software, Shanghai Jiao Tong University. His research interests include formal verification, programming language theory and computational learning theory.
Daniel W. Sun received his Ph.D. in Information Science from Japan Advanced Institute of Science and Technology (JAIST) in 2008. From 2008 to 2012, he was an assistant research manager in NEC central laboratories in Japan. From 2013, he has been working for National ICT Australia as a researcher. He is also a conjoint lecturer in School of Computer Science and Engineering, the University of New South Wales, Australia. His current research interests include big data, cloud computing, cybersecurity, system reliability, and data mining.
Xiaohong Wu received the Master and Bachelor degree in information engineering from Zhejiang University of Technology in 2004 and 1996. She is currently an associate professor in School of Information Engineering Huzhou University and a Ph.D. candidate in Shanghai University of Finance and Economics. Her current research interests include game theory, mechanism design and cloud computing.
Prem Prakash Jayaraman received the Ph.D. degree in computer science from Monash Univeristy, Melbourne, Australia, in 2011. He is currently a Research Fellow with RMIT University, Melbourne, Vic., Australia. He was a Postdoctoral Research Fellow with CSIRO Digital Productivity Flagship, North Ryde N.S.W., Australia, from 2012 to 2015. Prior to that, he worked as a Research Fellow and Lecturer with the Centre for Distributed Systems and Software Engineering, Monash University, Melbourne, Vic., Australia. His research interests include Internet of Things, cloud computing, mobile computing, sensor network middleware, and semantic Internet of Things.
Rajiv Ranjan received the PhD degree. He has been a reader (associate professor) of computing science at Newcastle University since September 1, 2015. He is an internationally renowned researcher in the areas of cloud computing, Internet of Things (IoT), and big data. By applying ground-breaking combination of well-founded formal models from four domains (Operations Research, Computational Statistics, Peer-to-Peer Networking, and Performance Engineering) of computer science, he has developed novel algorithmic techniques and distributed system architectures that facilitate service level agreement (SLA) driven autonomic provisioning of multimedia (e.g., content delivery networks), eScience (e.g., scientific work- flows), and IoT big data applications (e.g., remote sensing, smart homes, smart cities, etc.) applications over multiple private and public cloud data centres. He has authored approximately 150 scientific publications, including publications in the IEEE Transactions on Parallel and Distributed Systems (ERA A*), IEEE Transactions on Computers (ERA A*), Journal of Computer and System Sciences (ERA A*), and IEEE/ ACM World Wide Web conference (CORE A+). He is widely recognised by his peers through citations (5,500+ Google Scholar citations—https://goo.gl/7FONZN and 730+ Web of Science citations—http://goo.gl/St567J).