Elsevier

Computer Networks

Volume 74, Part B, 9 December 2014, Pages 22-33
Computer Networks

Energy-efficient and network-aware offloading algorithm for mobile cloud computing

https://doi.org/10.1016/j.comnet.2014.06.020Get rights and content

Abstract

We propose a new system architecture for mobile cloud computing (MCC) that includes a middle layer sitting between mobile devices and their cloud infrastructure or clones. This middle layer is composed of cloudlets and is thus called a cloudlet layer. Cloudlets are deployed next to IEEE 802.11 access points and serve as a localized service point in close proximity to mobile devices to improve the performance of mobile cloud services. On top of this new architecture, an offloading algorithm is proposed with the main aim of deciding whether to offload to a clone or a cloudlet. The decision-making takes into consideration the energy consumption for task execution and the network status while satisfying certain task response time constraints. We also introduce a data caching mechanism at cloudlets to further improve the overall MCC performance. Simulation results demonstrate the effectiveness and efficiency of the proposed system architecture and offloading algorithm in terms of response time and energy consumption.

Introduction

The past few years have witnessed a rapid shift in computing from the desktop to the cloud. To keep pace with advances in both wireless network technologies and mobile smart phones, there is an increasing need for provision of cloud services to mobile users via mobile wireless networks. This new research field is called mobile cloud computing (MCC) [1], [2], [3], [4], [5]. Because mobile devices, even modern smart phones, are constrained in size and weight, their resources for computation and communication are limited compared to their desktop counterparts [6]. Therefore, it makes sense to offload heavy mobile applications to more powerful machines in the cloud. Computing and service delivery are possible because of the advanced sensors built into most mobile phones currently on the market; these sensors include accelerometers, magnetometers, GPS chips, gyroscopes, and pressure sensors. The more sensors a device has, the more data need to be analyzed in various domains at the same time, which accentuates the need for more computational power. One critical issue in MCC is how battery power for mobile devices can be spared [7]. One effective approach is to offload some tasks from the mobile device to a remote cloud server for execution. This can also potentially reduce the task execution time because of the power of cloud servers. Kumar and Lu investigated the power consumption of mobile devices, including whether offloading can increase battery life [8]. Offloading should only occur when it is beneficial to the mobile application. Thus, tasks should only be offloaded to the cloud if the sum of the data transmission cost and the energy cost is smaller than when the tasks are executed locally on the mobile device.

We contribute to the exciting MCC research field by investigating a key problem for mobile application offloading. There is much work being carried out in this area, which largely falls into two categories: (1) task partitioning, which involves dividing an application into offloadable partition(s) and a local partition [2]; and (2) virtual machine (VM) selection [9], which involves choosing an appropriate VM to which to offload a partition. VMs are the key component of a cloud and they provide virtual resources such as CPU, memory, storage, and network interfaces in the same way as physical resources do. One common element of this research is the assumption that there is a perfect network connection and sufficient bandwidth between a mobile application on a mobile device and the remote cloud VM. This may not be an unrealistic assumption for wired networks, for which network bandwidth is usually abundant or at least not scarce. However, this is not the case for wireless networks [10], for which network bandwidth is not as abundant and a network connection may sometimes not even be available; this is more of a problem for mobile cellular networks such as 3G.

To address this issue, we use a new network-awareness perspective: in addition to considering network status parameters such as bandwidth and delays, we also take network types into account. We consider two types of wireless network in this study: IEEE 802.11 (i.e., WiFi) and mobile cellular networks such as 3G and LTE. These are the two mainstream wireless technologies that people interact with on a daily basis. Considering one without the other is not realistic and leads to a situation in which a mobile cloud system is not as efficient as it should be. This consideration is reflected in our proposed MCC system architecture, in which a middle layer is introduced between mobile devices and their corresponding cloud clones. This middle layer, called the cloudlet layer, is deployed next to WiFi access points (APs) in the proximity of mobile devices. The aim is to run clone equivalents named cloudlets in this layer, as illustrated in Fig. 1. The benefits of this approach are twofold. First, it can take advantage of the higher bandwidth of WiFi and switch the network connection from 3G to WiFi. Second, data caching can be carried out on cloudlets to some extent for applications such as data downloading from the Internet. 3G networks are typically proprietary and are not open to MCC service providers, whereas WiFi networks can be easily deployed. Otherwise, MCC service providers can relatively easily install their offloading software on the home servers of end users or on gateways next to their home APs. Our proposed MCC infrastructure is intended for MCC service providers rather than network providers or operators. As a result, our proposed offloading algorithm focuses on the offloading location (clone or cloudlet) when offloading a mobile application. By contrast, most of the current literature on offloading focuses on whether to offload a mobile application or not. Cloudlets can also be placed next to cellular base stations, as illustrated by the blue line (connecting a mobile device, cloudlet, and clone) in Fig. 1. However, this issue is beyond the scope of the present study and will be investigated in future work.

In summary, we propose a new MCC system architecture that contains a cloudlet middle layer above the existing cloud server infrastructure. We also propose an offloading algorithm that decides on where to offload a given mobile application. The objectives of the offloading algorithm are twofold: (1) to minimize the service response time, which is a direct factor of the quality of experience of cloud end users and (2) to save the battery life of mobile devices.

In comparison to the existing literature on MCC offloading, we make the following two major contributions:

  • We propose a new MCC system architecture with a cloudlet layer sitting between mobile devices and their traditional cloud infrastructure or clones. Cloudlets are deployed next to IEEE 802.11 APs and serve as a localized service point in close proximity to mobile devices, which improves the performance of MCC services.

  • On top of this new architecture, an offloading algorithm decides whether to offload to a clone or a cloudlet. The decision-making takes into consideration the energy consumption for task execution and the network status, while satisfying certain task response time constraints. Note that the energy efficiency considered here is from the perspective of mobile devices rather than the cloud servers.

The remainder of the paper is organized as follows. Related work is described in Section 2. Section 3 presents the proposed MCC system architecture, detailing each component and their relationships. Section 4 describes the proposed energy-efficient and network-aware offloading algorithm and an energy model in the context of our proposed MCC system architecture, and the two types of wireless network considered. The simulation results and performance analysis are presented in Section 5. The paper concludes with Section 6.

Section snippets

Related work

Our literature review focuses on the two main contribution areas: MCC system architecture and offloading algorithms. Since the core is offloading algorithms the discussion on MCC system architecture is from the perspective of offloading support.

Proposed offloading architecture

In this section, we describe user issues for conventional offloading architectures and propose an architecture with solutions that overcome these issues. We also validate the solutions.

Dynamic offloading algorithm

The offloading algorithm embedded in the remote execution decision engine decides on Whether and Where to offload data. The novelty of this algorithm is that it considers more than one offloading location as a parameter when deciding Where to offload. By contrast, conventional offloading algorithms support just one offloading location [34]. In the remainder of the paper, we assume that cloudlets are only located next to WiFi APs and mobile devices access clones using their cellular network via

Performance evaluation and analysis

The name Cloudlet + Clone is used hereafter to refer to our proposed architecture. The terms CloudletClone and CloneOnly in figures refer to Cloudlet + Clone and a conventional architecture, respectively.

Conclusions and future work

We proposed a new MCC system architecture called Cloudlet + Clone, containing a new middle layer called a cloudlet layer. This cloudlet layer sits between mobile devices and their traditional cloud infrastructure or clones. Cloudlets are deployed next to WiFi APs and serve as a localized service point in close proximity to mobile devices to improve the performance of mobile cloud services in terms of response time. An offloading algorithm for deciding Whether and Where to offload is applied on

Acknowledgement

The work reported here was partially funded by the EU FP7 Projects MONICA (GA-2011-295222), CLIMBER (GA-2012-318939), and EVANS (GA-2010-269323).

Chathura Magurawalage received his B.Sc. Hons. degree in Computer Science from the University Of Essex, UK. He Joined the Network Convergence Laboratory at the University of Essex in 2012, after being awarded the University Of Essex scholarship to carry out research as a PhD student. His current research interests include Mobile Cloud Computing, Cloud computing and Computer Networks.

References (43)

  • S. Qureshi, T. Ahmad, K. Rafique, S. ul Islam, Mobile cloud computing as future for mobile applications –...
  • L. Yang et al.

    A framework for partitioning and execution of data stream applications in mobile cloud computing

  • B.-G. Chun et al.

    Clonecloud: elastic execution between mobile device and cloud

  • S. Kosta, A. Aucinas, P. Hui, R. Mortier, X. Zhang, Thinkair: dynamic resource allocation and parallel execution in the...
  • D. Chang, G. Xu, L. Hu, K. Yang, A network-aware virtual machine placement algorithm in mobile cloud computing...
  • IEEE

    Standard for local and metropolitan area networks part 16: air interface for fixed and mobile broadband wireless access systems amendment 2: physical and medium access control layers for combined fixed and mobile operation in licensed bands and corrigendum 1

    (2006)
  • J.D. Power and Associates, US Wireless Smartphone and Traditional Mobile Phone Satisfaction Studies,...
  • K. Kumar et al.

    Cloud computing for mobile users: can offloading computation save energy?

    Computer

    (2010)
  • H. Wu, Q. Wang, K. Wolter, Methods of cloud-path selection for offloading in mobile cloud computing systems, in: IEEE...
  • Y. Wu et al.

    A new analytical model for multi-hop cognitive radio networks

    IEEE Trans. Wireless Commun.

    (2012)
  • R. Kemp et al.

    Cuckoo: a computation offloading framework for smartphones

  • G.F. Nan et al.

    Distributed resource allocation in cloud-based wireless multimedia social networks

    IEEE Netw. Magaz.

    (2014)
  • S. Kosta, C. Perta, J. Stefa, P. Hui, A. Mei, Clone2clone (c2c): Enable Peer-to-Peer Networking of Smartphones on the...
  • E. Cuervo et al.

    Maui: making smartphones last longer with code offload

  • K.K. Rachuri et al.

    Sociablesense: exploring the trade-offs of adaptive sampling and computation offloading for social sensing

  • M.V. Barbera, S. Kosta, A. Mei, V.C. Perta, J. Stefa, CDroid: Towards a cloud-integrated mobile operating system, in:...
  • M. Satyanarayanan et al.

    The case for VM-based cloudlets in mobile computing

    IEEE Pervasive Comput.

    (2009)
  • R. Yu et al.

    Toward cloud-based vehicular networks with efficient resource management

    IEEE Netw.

    (2013)
  • D.T. Hoang, D. Niyato, P. Wang, Optimal admission control policy for mobile cloud computing hotspot with cloudlet, in:...
  • D. Niyato, P. Wang, E. Hossain, W. Saad, Z. Han, Game theoretic modeling of cooperation among service providers in...
  • A.-D. Nguyen, P. Senac, V. Ramiro, How mobility increases mobile cloud computing processing capacity, in: 1st...
  • Cited by (72)

    • Makespan-minimization workflow scheduling for complex networks with social groups in edge computing

      2020, Journal of Systems Architecture
      Citation Excerpt :

      The workflow of a mobile application is generally represented by a DAG. Recently, scheduling workflows in the above-mentioned device-to-device-enabled edge computing has drawn more and more attention [17–22]. Most existing algorithms are based on the assumption that the edge computing environment has a single all-to-all network, i.e., each device can communicate with all other devices in this network.

    • Edge computing: A systematic mapping study

      2023, Concurrency and Computation: Practice and Experience
    • Machine Learning Approach for Intelligent and Sustainable Smart Healthcare in Cloud-Centric IoT

      2023, International Journal of Intelligent Systems and Applications in Engineering
    • A Context Sensitive with Effective Task Migration in Mobile Cloud Computing Services

      2023, 2023 3rd International Conference on Computing and Information Technology, ICCIT 2023
    View all citing articles on Scopus

    Chathura Magurawalage received his B.Sc. Hons. degree in Computer Science from the University Of Essex, UK. He Joined the Network Convergence Laboratory at the University of Essex in 2012, after being awarded the University Of Essex scholarship to carry out research as a PhD student. His current research interests include Mobile Cloud Computing, Cloud computing and Computer Networks.

    Kun Yang received his Ph.D. from the Department of Electronic & Electrical Engineering of University College London (UCL), UK. He is currently a full Professor and the Head of Network Convergence Laboratory (NCL) in the School of Computer Science and Electronic Engineering, University of Essex, UK. Before joining in University of Essex at 2003, he worked at UCL on several European Union research projects in the area of IP network management and service engineering. His current major research interests include wireless networks, heterogeneous wireless networks, fixed mobile convergence, future Internet technologies such as information centric networking, and cloud computing. He has published more than 150 journal and conference papers in the above areas. He serves on the editorial boards of both IEEE and non-IEEE journals. He is a Senior Member of IEEE, a Fellow of IET.

    Liang Hu. Received his M.S. and Ph.D. degrees in Computer Science from Jilin University, in 1993 and 1999 respectively. Currently, he is a professor and doctoral supervisor at the College of Computer Science and Technology, Jilin University, China. His research areas are Network Security and Distributed Computing, including related theories, models and algorithms of PKI/IBE, IDS/IPS, and Parallel Computing. He has led a lot of research projects, and has published 100+ journal papers. He is a member of the China Computer Federation.

    Jianming Zhang received his Ph.D. degree in 2010 from Hunan University, China. He received his M.S. and B.E. in 2001 and 1996, respectively, from the National University of Defense Technology and Zhejiang University, China. Currently, he is an associate professor in the School of Computer and Communication Engineering at Changsha University of Science and Technology, China. His main research interests lie in the areas of wireless multimedia sensor networks, pattern recognition and computer vision. He has served as an invited member of the technical program committee of several international conferences such as MobiQuitous 2013, mCloud 2013 and CloudID 2013. He is a member of ACM and CCF.

    View full text