Cognitive Mobile Cloud Computing

Originally posted in Sec-IG blog (link to the original post)

The huge growth in mobile data traffic is generally because of increasingly sophisticated smart phone applications (e.g. Computer Vision, Augmented Reality, 3D gaming, cameras, and security). These applications are a major market driver for mobile equipment providers, service providers and IT players. The increasing sophistication of these applications, combined with the resource constraint on the hand-helds running multiple apps and user preference for lighter mobile devices (e.g. iPhone7) and wearables (e.g. Microsoft HoloLens, Apple watch, Fitbit) has made a case for offloading some of the computations supporting the apps to a resource rich cloud.

The term cloud offloading can mean either data flow offloading in networking applications or offloading computation intense processes on to the cloud. Here, we refer to the latter. Cloud offloading can be classified into three categories: (a) those that always offload to the cloud; (b) “all or nothing offloading” where either the entire application is offloaded to the cloud or executed locally, typically using an energy threshold to decide between offloading and not [1]; and (c) piecewise decisions, where some parts are executed locally while the others are offloaded to the cloud [2]. The third category offers the most flexibility for trade-offs, and can be done either at the coarse component level or at finer, method or instruction levels.

This computation offloading, especially for the apps consuming large data, occupies a considerable part of limited bandwidth over wireless networks. Spectrum-aware cognitive mobile cloud computing is the new concept (started from 2015 [3]) to use dynamic spectrum opportunistically from wireless networking to effect computation offloading and scheduling solutions that achieves efficient trade-offs between the mobile device and wireless resources. In this concept, cognitive cloud offloader schedules appropriate components of the application to run either on the mobile device or on the resource-rich cloud, while staying adaptive to the realtime changes of the wireless network [4]. Here, all the viable radio available interfaces (e.g. WiFi, LTE) in multi-RAT enabled devices are deployed to increase the throughput capacity and reliability, while balancing energy consumption for data transferring. Moreover, this concept allows for more degrees of freedom in the solution by moving away from a compiler pre-determined scheduling order for the app component tasks towards a more spectrum aware scheduling order. Hence, this solution can shorten runtimes by parallel processing proper app component tasks in the mobile device and the cloud [5]. Finally, the dynamic algorithms used in the cognitive cloud offloader have several control knobs that can be used to balance the trade-off among the relative importance of battery power, end-to-end network delay, bandwidth, monetary cost of network and cloud access.


Figure 1. The Future: cognitive cloud offloading in heterogeneous mobile networks [7].

Cognitive cloud offloader can be used in the new wireless technologies such as mobile edge computing (MEC) and C-RAN as shown in Figure 1. With the advent of newer mobile edge computing (MEC) paradigms [6], where the cloud is closer to the end device than before, the offloading option becomes even more attractive in terms of keeping up with near real-time user responsiveness of the apps.

Given recent advances in technologies that enable bandwidth aggregation in wireless devices, the cognitive cloud offloader solution is implementable in practice. Other works that fall under general umbrella of the radio-aware computation offloading include, where the best of the available wireless interfaces is chosen (only one of the wireless interfaces) for data transfer, rather than a solution that considers using all of the radio interfaces simultaneously [8,9]. In recent works cloud offloading scheduling mechanisms are proposed for queue stability, but these works only deal with multi-channel systems, not multi-radio networks [1]. Evaluation results in [3,4,6] for the cognitive cloud offloader, using real data measurements from Android smart phones running multi-component applications and Amazon EC2 and NSFCloud over LTE and WiFi, show that the proposed strategies reduce energy consumption for the mobile device by 23% to 68% and speed-up in runtime by 46% to 66% compared to the state of art. Note also that the proposed suite of solutions provide higher degrees of freedom in protocol design for serving applications over multi-RAT enabled mobile devices and is an inter-disciplinary solution combining mobile computing and cognitive radio networking.


[1] W. Zhang, Y. Wen, K. Guan, D. Kilper, H. Luo, and D. Wu, “Energy-optimal mobile cloud computing under stochastic wireless channel,” IEEE Transactions on Wireless Communications, vol. 12, no. 9, pp. 4569–4581, Sep. 2013.

[2] E. Cuervo, A. Balasubramanian, D.-k. Cho, A. Wolman, S. Saroiu, R. Chandra, and P. Bahl, “MAUI: Making smartphones last longer with code offload,” in Proceedings of the International Conference on Mobile Systems, Applications, and Services, ser. MobiSys. ACM, 2010, pp. 49–62.

[3] “System Apparatus and Methods for Cognitive Cloud Offloading in a Multi-RAT Enabled Wireless Device”, S. Eman Mahmoodi and K. P.Subbalakshmi, Provisional U.S. Patent filed, US 62/262,624, December 2015.

[4] S. E. Mahmoodi and K. P. S. Subbalakshmi, “A Time-Adaptive Heuristic for Cognitive Cloud Offloading in Multi-RAT Enabled Wireless Devices,” in IEEE Transactions on Cognitive Communications and Networking, vol. 2, no. 2, pp. 194-207, June 2016.

[5] S. E. Mahmoodi, R. N. Uma, and K. P. Subbalakshmi, “Optimal joint scheduling and cloud offloading for mobile applications,” IEEE Transactions on Cloud Computing, vol. PP, no. 99, pp. 1–1, early access 2016.

[6] M. Patel, B. Naughton, C. Chan, N. Sprecher, S. Abeta, A. Neal et al., “Mobile-edge computing introductory technical white paper,” White Paper, Mobile-edge Computing (MEC) industry initiative, 2014.

[7] S. Eman Mahmoodi, K. P. Subbalakshmi, and R. N. Uma. 2016. Harnessing spectrum awareness to enhance mobile computing: poster. In Proceedings of the 22nd Annual International Conference on Mobile Computing and Networking (MobiCom ’16). ACM, New York, NY, USA, 460-461.

[8] D. Huang, P. Wang, and D. Niyato, “A dynamic offloading algorithm for mobile computing,” IEEE Transactions on Wireless Communications, vol. 11, no. 6, pp. 1991–1995, Jun. 2012.

[9] P. Shu, F. Liu, H. Jin, M. Chen, F. Wen, and Y. Qu, “eTime: Energyefficient transmission between cloud and mobile devices,” in IEEE Proceedings of INFOCOM, 2013, pp. 195–199.


S. Eman Mahmoodi
The Information Networks and Security Lab
Department of Electrical & Computer Engineering,
Stevens Institute of Technology, Hoboken, NJ, USA

S. Eman Mahmooodi is currently pursuing his PhD degree at the Department of Electrical and Computer Engineering, Stevens Institute of Technology. He received the BS and MS degree in Electrical Engineering from Iran University of Science and Technology, respectively in 2009 and 2012. He has been working on Mobile Cloud Computing, Optimization Algorithms and Predictive Modeling, Internet of Things, Cognitive Networking, and Wireless Communications. Mahmoodi is a Stevens Innovation and Entrepreneurship Doctoral Fellow, and he has also received the graduate student inventor award from the New Jersey Inventors Hall of Fame (NJIHoF), 2016.

Leave a Reply

Your email address will not be published.

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>