MAGA: A MOBILITY-AWARE COMPUTATION OFFLOADING DECISION FOR DISTRIBUTED MOBILE CLOUD COMPUTING
Fail computation offloading requests. Latest researches on human mobility show that mobility of users present inherent patterns, periodicity and predictability. This motivates Distributed mobile cloud computing (MCC) is the new paradigm for providing ubiquitous cloud resources to mobile users with low latency. Mobility is an important factor in distributed mobile cloud computing which may incur intermittent connectivity and consequently us to propose a mobile access prediction algorithm based on tail matching sub-sequence, whose effectiveness and accuracy is validated by experiments using reality mobility dataset. Then MAGA, a mobility-aware offloading decision method for distributed mobile cloud computing is proposed in this paper for single-job, multi-component and multi-site offloading scenario. The proposed mobile access prediction is used in MAGA for cloudlet reliability estimation. An integer encoding based adaptive genetic algorithm is used for offloading decision. Experiment results show the performance advantages of MAGA.
Index Terms—Offloading decision, mobility, cloudlet, distributed mobile cloud computing, genetic algorithm
MOBILE cloud computing (MCC) emerged as a new research domain to augment various mobile devices by leveraging heterogeneous cloud resources. MCC integrates the cloud computing into the mobile environment and overcomes obstacles related to resource limitations (e.g., battery life, storage, CPU processing capacity), environment (e.g., heterogeneity, scalability, and availability), and security (e.g., reliability and privacy) discussed in mobile computing. However, many MCC solutions are proposed based on infrastructure clouds, which are typically far from the mobile users. Therefore, the high WAN (Wide Area Network) latency makes those solutions insufficient for real-time applications (e.g., augment reality applications). In recent years, new efforts have been made by researchers to push the resources (computation, communication, control and storage) closer to the end users, i.e., MCC evolves from centralized architecture to distributed architecture. Cloudlets, mobile edge computing (MEC) and fog computin concepts were proposed as different paradigms of distributed mobile cloud computing. In distributed mobile cloud computing, computation offloading is one of the critical technologies to improve the performance and reduce local execution cost through migrating heavy computation tasks to powerful servers in clouds. However, offloading is not always beneficial for the mobile user because the time, energy and communication overhead incurred by offloading. Therefore, offloading decision plays an important role to balance the benefits and overheads of offloading.
A number of researches have devoted to offloading decision in mobile cloud computing, which determine whether and what computation to migrate for improving performance and saving energy . Those offloading strategies could be classified into static offloading strategy and dynamic offloading strategy. The offloading decision is made statically during program development in static strategies, and is made dynamically during execution in dynamic ones. Although static strategies showed lower overhead, they are not adaptive to the dynamically changed environments of MCC. Conversely, dynamic strategies are capable of improving offloading performance due to they can vary depending on the users’ context. The offloading could be conducted for a single user or for multiple users. Single user offloading strategies aim to optimize the offloading QoS (Quality of Service) from the single user perspective, such as to decrease the mobile device energy consumption or to improve the offloading efficiency. The multi-user offloading strategies often accommodate the whole cloud system efficiency in energy or resource utilization. In addition, Some offloading strategies are conducted for a whole application , and some others are conducted at the granularity of service components.
Mobility Consideration in Offloading
As mentioned in Section, mobility is an un-neglected key attribute in MCC. Mobility introduces dynamic network environments, may incur intermittent connectivity that can fail user requests. Some recent works have recognized mobility as an important factor in offloading decision. Reference developed an offloading algorithm for intermittently connected cloudlet system. User mobility is taken into consideration to evaluate the successful offloading probability. But in , the user is assumed to move straight in the cloudlet coverage, and the centrifugal direction or centripetal direction of user movement is used to estimate the cloudlet availability. Reference also considered user mobility in MCC resource allocation. But users’ mobility is simple assumed as random mobility. Reference designed a robust offloading decision system for service workflow. RWP (Random Waypoint Mobility) model is assumed for user mobility, and is involved into the proposed strategy to consider users’ locations during service execution. Reference proposed an offloading decision method in which a Markov model is used as user mobility model for mobility prediction. Analysis on the above works about mobility aware offloading decision shows that the mobility models assumed in these works are relatively simple (e.g., straight movement, RWP model, simple random mobility), so that they fail to portray the reality and complex user mobility characteristics, such as the inherent mobility patterns, the periodicity etc. Such observations motivate us to propose MAGA, which is expected to utilize the inherent user mobility patterns to optimize offloading decision.
This paper focused on offloading decision in distributed mobile cloud computing, aiming at improving offloading success rate and decrease energy consumption at mobile devices when satisfying the job completion time requirement. MAGA, a mobility-aware computation offloading decision algorithm was proposed for the single-user, multi-component and multi-cloudlet scenario. The first contribution of this paper was to propose a TMSS (Tail Matching Sub-Sequence) based mobile access prediction method. It was validated to be effective and can provide satisfactory prediction accuracy because it explored and utilized the inherent mobility patterns and periodicity feature of human mobility. Then this prediction method was used in MAGA for cloudlet reliability evaluation. The second contribution of this paper was to propose an improved genetic algorithm for optimal offloading decision. Integer encoding was employed to well fit the multi-component and multi-cloudlet scenario. Adaptive reproduction operations (i.e., crossover and mutation) were adopted to improve algorithm efficiency. Experiment results showed that MAGA could improve offloading success rate, decrease energy consumption at mobile devices, and also presented better convergence and global searching capability than canonical genetic algorithm.
. Z Sanaei, S Abolfazli, A Gani et al., “Heterogeneity in Mobile Cloud Computing: Taxonomy and Open Challenges,” IEEE Communications Surveys & Tutorials, vol. 16, no. 1, May 2014, pp. 369-392.
. X. Chen, “Decentralized Computation Offloading Game for Mobile Cloud Computing,” IEEE Transactions on Parallel and Distributed Systems, vol. 26, no. 4, April 2015, pp. 974-983.
. Y. Cui, J. Song, K. Ren, M. Li, Z. Li, Q. Ren, et al., “Software defined cooperative offloading for mobile cloudlets,” IEEE/ACM Transactions on Networking, pp. 1-15, DOI: 10.1109/TNET.2017.2650964.
 H. T. Dinh, C. Lee, D. Niyato, P. Wang, “A survey of mobile cloud computing: architecture, applications, and approaches,” Wireless Communications & Mobile Computing, vol. 13, no. 18, 2013, pp. 1587–1611.
. Y. Li, W. Gao, “Minimizing Context Migration in Mobile Code Offload,” IEEE Transactions on Mobile Computing, vol. 16, no. 4, 2017, pp. 1005-1018.
. S. Guo, B. Xiao, Y. Yang et al., “Energy-efficient dynamic offloading and resource scheduling in mobile cloud computing,” Proceedings of IEEE INFOCOM 2016, San Francisco, CA, USA, April 10-14, 2016. pp. 1-9.
. T. Verbelen, P. Simoens, F. D. Turck, et al., “Cloudlets: bringing the cloud to the mobile user,” ACM Workshop on Mobile Cloud Computing & Services, 2012. pp. 29-36.
. M. Satyanarayanan, P. Bahl, R. Caceres, and N. Davies, “The case for VM-based cloudlets in mobile computing,” IEEE Pervasive Computing, vol. 8, no. 4, Oct. 2009, pp. 14–23.
. A. Ahmed, E. Ahmed, “A survey on Mobile Edge Computing,” 10th IEEE International Conference on Intelligent Systems and Control, Coimbatore, India, Jan. 7-8, 2016. pp. 1-8.
. F. Bonomi, R. Milito, J. Zhu, S. Addepalli, “Fog Computing and Its Role in the Internet of Things,”, Edition of the Mcc Workshop on Mobile Cloud Computing ACM, Helsinki, Finland, August 17, 2012. pp. 13-16.
. M. Chiang, T. Zhang, “Fog and IoT: An Overview of Research Opportunities. IEEE Internet of Things Journal. VOL. 3, NO. 6, DECEMBER 2016. pp. 854-864.
. Y. Zhang, D. Niyato, P. Wang, “Offloading in Mobile Cloudlet Systems with Intermittent Connectivity,” IEEE Transactions on Mobile Computing, Vol. 14, No. 12, December 2015. pp. 2516-2529