On the throughput-energy tradeoff for data transmission between cloud and mobile devices

doi:10.1016/j.ins.2014.06.022

Information Sciences

Volume 283, 1 November 2014, Pages 79-93

https://doi.org/10.1016/j.ins.2014.06.022 Get rights and content

Abstract

Mobile cloud computing has recently emerged as a new computing paradigm promising to improve the capabilities of resource-constrained mobile devices. As the data processing and storage are moved from mobile devices to powerful cloud platforms, data transmission has become an important issue affecting user experiences of mobile applications. One of the challenges is how to optimize the tradeoff between system throughput and energy consumption, which are potentially conflicting objectives. Inspired by the feasibility of transmission scheduling for prefetching-friendly or delay-tolerant applications, we mathematically formulate this problem as a stochastic optimization problem, and design an online control algorithm to balance such an energy-performance tradeoff based on the Lyapunov optimization framework. Our algorithm is able to independently and simultaneously make control decisions on admission and transmission to maximize a joint utility of the average application throughput and energy cost, without requiring any statistical information of traffic arrivals and link bandwidth. Rigorous analysis and extensive simulations have demonstrated both the system stability and the utility optimality achieved by our control algorithm.

Introduction

Mobile devices, such as smartphone, PDA, and tablet-PC, are gradually becoming an important part of human life as the most convenient communication tools unbounded by time and space. However, mobile devices are facing severe challenges in resources (e.g., computation, storage and energy) and communications (e.g., bandwidth and mobility), which greatly impede the improvement of user experiences. In recent years, the paradigm of mobile cloud computing has been introduced to extend capabilities of mobile devices, by taking advantage of high-speed wireless communications and high-performance cloud platforms [7], [21] to help gather, store and process data for the mobile devices [22], [14]. In this new paradigm, the cloud-based mobile applications usually require frequent data exchanges between mobile devices and cloud platforms [44]. With the surging popularity of mobile cloud computing, it has been found that traffic from mobile devices grows three times faster than that from wired devices, and will exceed the latter by year 2016 [8]. Unlike wired devices, current mobile devices and their daily use are seriously constrained by the energy capacity of batteries [23], [17]. Hence, the data transmission strategy for mobile devices has to take into account not only providing sufficient system throughput to satisfy application requirements, but also conserving precious battery energy to prolong operational lifetime.

This paper explores such a throughput-energy trade-off for prefetching-friendly or delay-tolerant applications on mobile devices having multiple types of network interfaces. To support pervasive Internet access, current mobile devices are increasingly being equipped with more than one wireless interfaces [9], such as WiFi, 3G-HSDPA, and 4G-LTE [3], [13], that have substantially different characteristics. Firstly, the availability of these types of networks can vary significantly at different places [32], [35]. At least as of now, the coverage and penetration capabilities of 3G and 4G are much higher than those of WiFi. Secondly, the achievable data rates of these interfaces can vary considerably over time, and are sometimes far less than the nominal values specified by the technical standards [13], [35]. For example, it is reported in [13] that the downlink bandwidth measured in U.S. range from 0.35Mbps to 19.27 Mbps for 802.11g WiFi (up to 54 Mbps), and from 2.08 Mbps to 30.80 Mbps for LTE (up to 100 Mbps). Thirdly, the energy costs for transmitting a given amount of data over these wireless interfaces could differ by an order of magnitude [13], [30]. It has been revealed by empirical studies in [13] that the radio power level has a linear relationship with link transmission bandwidth, while the power coefficients are distinct for different types of interfaces. Based on above-mentioned reasons, the selection of available wireless interfaces for data transmission has direct impacts on system throughput and energy consumption of mobile devices.

Fortunately, many of the mobile applications are naturally prefetching-friendly or delay-tolerant, to different degrees, so that it is possible to switch among multiple radio interfaces to achieve energy-efficient data transfer. For example, mobile users may rely on the online map service when visiting new places. When they use the mobile client to request online maps, the cloud could pre-push some potentially useful maps (e.g., historic sites) to the device during periods of good connectivity of a lower-energy link. On the other hand, delay-tolerant applications such as online social networking can be set to fetch the content update at specific intervals. This create opportunities for energy saving by deferring the transmission of the latest data until a satisfactory link connection becomes available. However, one-sided pursuit of energy saving may degrade system throughput and application performance.

To address the challenges above, we propose a new online control algorithm, MOTET, for Mobile device Optimization on the Throughput-Energy Tradeoff using the Lyapunov optimization framework [27]. MOTET maximizes a joint utility of the sum throughput of applications and the energy costs of device, by independently and simultaneously making online decisions to control admission and transmission behaviors [11], [20]. It is associated with two control parameters, i.e., throughput-energy parameter and stability-utility parameter, which can be appropriately tuned to provide a desired performance tradeoff among application throughput, energy cost and service delay. Specifically, a non-negative parameter $θ$ is used to normalize the values of throughput and energy to make them comparable in the utility function. Meanwhile, MONET can obtain a time-average utility within a deviation of $O (1 / V)$ from optimality, while bounding the traffic queue length and the traffic service delay by $O (V)$ , where V is a non-negative control parameter representing a design knob of the stability-utility tradeoff (i.e., how much we emphasize utility maximization compared to system stability). MOTET operates without requiring any statistical knowledge of traffic arrivals and link conditions, and is computationally efficient for implementing on resource-constrained mobile devices. We thoroughly analyze the performance of our new proposed online control algorithm with rigorous theoretical analysis. To complement the analysis, we conduct a simulation study to evaluate MOTET using datasets from real-world measurements on wireless link bandwidth and transmission energy consumption of mobile devices [13]. Experimental results demonstrate that MOTET can approach a time-average utility that is arbitrarily close to the optimum, while still maintaining strong stability and low congestion. Furthermore, with an appropriate throughput-energy parameter, MOTET can significantly reduce the energy expenditure while only incurring a marginal sacrifice in the system throughput. To our knowledge, prior work has not explored such a tradeoff issue on mobile devices, and our use of the Lyapunov optimization framework for solving this issue is also novel.

The rest of this paper is organized as follows: Section 2 reviews some related work. Section 3 describes the theoretical model for throughput-energy tradeoff, and also formulates the objective problem. Section 4 presents the optimal control algorithm, and Section 5 provides an analysis on performance bounds of our algorithm. Section 6 shows the performance evaluation results. Finally, Section 7 concludes the paper.

Section snippets

Related work

The contribution of our work lies in the intersection of the following two important cutting-edge research topics.

Basic throughput-energy tradeoff model

As shown in Fig. 1, we consider a mobile device user who has M heterogeneous applications $m \in {0, 1, 2, \dots, M}$ running on a cloud platform [34]. The whole system operates in discrete time with unit time slots $t \in {0, 1, 2, \dots}$ . The data (i.e., content) generated by each application is processed in a corresponding queue, denoted as $Q (t) ≜ (Q_{1} (t), \dots, Q_{M} (t))$ , in which $Q_{m} (t)$ represents the queue backlog of application m’s data to be transmitted from the cloud to the mobile device at the beginning of time slot t. In

Problem transformation

To solve problem (6), we need to transform it to a solvable form. The Lyapunov optimization framework has provided the “auxiliary variables” to transform the problem into a new one involves only time averages rather than functions of time averages, and hence can be solved by the drift-plus-penalty framework [27]. Since the utility function $g (r_{m})$ is concave and non-linear, we introduce auxiliary variables $γ_{m}$ for each admitted traffic stream $R_{m} (t), \forall m \in {1, \dots, M}$ , as follows: $\begin{matrix} \max & \sum_{m = 1}^{M} g (γ_{m}) - θ p \\ s . t . & γ_{m} ⩽ r_{m}, \in \end{matrix}$

Algorithm analysis

In this section, we analyze the performance bound of our MOTET algorithm.

Theorem 2

(Algorithm Performance) Implementing the MOTET algorithm in every time slot for any fixed control parameter $V ⩾ 0$ , yields the following performance bounds:(1) The worst case queue backlog for each queue $Q_{m}$ is upper bounded by a finite constant $Q_{m}^{\max}$ for all t: $Q_{m} (t) ⩽ Q_{m}^{\max} ≜ V + 2 A_{m}^{\max}$ (2) Assuming FIFO service, and $Q_{m} (t) ⩽ Q_{m}^{\max}, Z_{m} (t) ⩽ Z_{m}^{\max}$ for all $t \in {0, 1, 2, \dots}$ . Then, the worst-case delay for data in $Q_{m}$ is upper bounded by the

Performance evaluation

In this section, we evaluate the proposed MOTET algorithm using datasets from real-world measurements on wireless link bandwidth and transmission energy consumption of mobile devices.

Conclusion

To reduce energy consumption for cloud-based mobile applications and prolong operational lifetime for energy-constrained mobile devices, we design and analyze an optimal online control algorithm, MOTET, to balance the tradeoff between the system throughput and the energy consumption in mobile cloud scenarios. By applying rigorous Lyapunov optimization framework, MOTET is able to independently and simultaneously make the control decisions on traffic admission and data transmission for the mobile

Acknowledgment

We would like to thank Dr. Junxian Huang (University of Michigan) for providing us the UMICH dataset from 4GTest measurement [13]. We are also grateful to Dr. Yunlu Liu (China Mobile Research Institute) for useful discussions. This work was supported by the National Natural Science Foundation of China under Grants 61202430 and 61303245, the State Key Lab of Astronautical Dynamics of China under Grant 2014ADL-DW0401, and the Science and Technology Foundation of Beijing Jiaotong University under

References (44)

T. Amin et al.
Dynamic programming approach to optimization of approximate decision rules
Inform. Sci.
(2013)
P.-C. Chao et al.
Multi-agent-based cloud utilization for the it office-aid asset distribution chain: an empirical case study
Inform. Sci.
(2013)
A. Gani et al.
A review on interworking and mobility techniques for seamless connectivity in mobile cloud computing
J. Netw. Comput. Appl.
(2014)
S.-H. Lim et al.
Energy-aware optimal cache consistency level for mobile devices
Inform. Sci.
(2013)
J. Lloret et al.
Architecture and protocol for intercloud communication
Inform. Sci.
(2014)
L.A. Mozelli et al.
A systematic approach to improve multiple lyapunov function stability and stabilization conditions for fuzzy systems
Inform. Sci.
(2009)
W. Pedrycz et al.
Building the fundamentals of granular computing: a principle of justifiable granularity
Appl. Soft Comput.
(2013)
Y. Zhuang et al.
Efficient and robust large medical image retrieval in mobile cloud computing environment
Inform. Sci.
(2014)
G. Ananthanarayanan, V.N. Padmanabhan, L. Ravindranath, C.A. Thekkath, Combine: leveraging the power of wireless peers...
Apple, Apple – ipad – view all the technical specifications, Apple Inc., 2013....

AT&T, AT&T high speed internet access, AT&T Inc., 2013....

N. Balasubramanian, A. Balasubramanian, A. Venkataramani, Energy consumption in mobile phones: a measurement study and...

A. Chakraborty, S. Das, Adapp: an adaptive network selection framework for smartphone applications, in: Proceeding of...

Cisco, Cisco visual networking index: forecast and methodology, 2012–2017, Cisco Systems Inc., 2013....

Y. Guo et al.

Optimal power management of residential customers in the smart grid

IEEE Trans. Parallel Distrib. Syst.

(2012)

F. Gustavo et al.

Qos-oriented admission control in hsdpa networks

Netw. Protocols Algor.

(2009)

I.-H. Hou, P.R. Kumar, Utility-optimal scheduling in time-varying wireless networks with delay constraints, in:...

J. Huang, F. Qian, A. Gerber, Z.M. Mao, S. Sen, O. Spatscheck, A close examination of performance and power...

K. Kumar et al.

A survey of computation offloading for mobile systems

Mob. Netw. Appl.

(2013)

H.A. Lagar-Cavilla, K. Joshi, A. Varshavsky, J. Bickford, D. Parra, Traffic backfilling: subsidizing lunch for...

K. Lee et al.

Mobile data offloading: how much can wifi deliver?

IEEE/ACM Trans. Netw.

(2013)

F. Liu et al.

Gearing resource-poor mobile devices with powerful clouds: architectures, challenges, and applications

IEEE Wireless Commun.

(2013)

Cited by (77)

Creating Universum for class imbalance via locality and its application in multiview subspace learning
2023, Information Sciences
Since the existing methods for generating Universum fail in class imbalance cases, we propose a new method for creating Universum via locality and betweenness, named L-CIBU. Compared to existing methods for generating Universum, L-CIBU not only generates appropriate Universum in the case of class imbalance but also has higher efficiency. Given the existence of class imbalance in multiview learning, we introduce Universum to multiview subspace learning (MvSL) to further enhance its performance and L-CIBU happens to be able to overcome this problem. As far as we know, this is the first time to introduce Universum into MvSL with more than two views. Specifically, we introduce Universum created by L-CIBU into multiview discriminant analysis (MvDA) and propose Universum MvDA (UMvDA). By analyzing several randomly generated imbalanced datasets, it is intuitively demonstrated that L-CIBU is more effective than FIBU and CIBU in the case of class imbalance. The experimental results on two real datasets show that the time required for creating the Universum using L-CIBU is $25 % \sim 95 %$ less than that of FIBU and CIBU. Moreover, UMvDA achieves an accuracy improvement of approximately $1 % \sim 6 %$ compared to MvDA, Universum canonical correlation analysis (UCCA), and Universum discriminant canonical correlation analysis (UDCCA).
Meta-analytical comparison of energy consumed by two sorting algorithms
2022, Information Sciences
Citation Excerpt :
In the same manner, information accessed and processed by mobile phones has dramatically increased over the years. This dramatic growth unfortunately call for an increased power supply or alternatively highly optimised energy efficient algorithms [16]. Sorting algorithms are used to arrange elements in a specific order [17].
Mobile devices performance and uptime heavily depend on energy consumed at the hardware and software level. Hence implementation of efficient algorithms has become a crucial aspect for increasing the performance of such systems and battery life for mobile devices. Sorting algorithms are implicitly the building block of many program implementation. Over the past years, researchers have spent more time optimizing hardware components to reduce their energy consumption. However, it has not been so clear which sorting algorithm is more energy efficient. In this study, we conduct a meta-analytical comparison of the energy consumed by the two most common sorting algorithms namely quick sort and merge sort. Our study mainly focused on energy consumption for mobile devices and embedded systems. For our meta-analysis and literature review, we took into consideration studies published not more than 20 years ago. The meta-analytical results show that there is no significant difference between both algorithms in terms of energy efficiency.
Design and analysis of a decision intelligent system based on enzymatic numerical technology
2021, Information Sciences
Citation Excerpt :
Hence, enzymatic numerical P systems (ENPSs) were proposed as membrane computing systems in which enzyme-like variables allow the existence of different production functions for each membrane and let the P system be more flexible in calculating [13,14]. Although an ENPS is powerful in different applications [9], it still cannot form an effective theory to support the decisional evolution process [15,16], because its attributes and mechanisms are mainly designed to handle the numerical values. In many studies, decisional theory has shown its importance [17,18] and we want to combine membrane computing with decisional theory.
A P system is a finite, discrete and distributed model with a parallel-layered and net structure. The enzymatic numerical P system (ENPS) has enzyme-like variables that allow each membrane to host more than one production function, and it has been widely used in economics and controllers for autonomous mobile robots. Although the ENPS has shown powerful abilities in numerical calculation, as a decisional process, it had not been able to express its mechanism of evolution, and the existing P system has a limited ability to process the assigned cells in different evolutions. Furthermore, the present decision models have a limited ability to process large-scale decisional tasks. To set up the decisional P system theory and render the existing ENPS more flexible, we present a restriction enzyme to found a conditional enzymatic numerical P system. In this system, we present a series of decisional enzymes and rebuild the structures of cells to achieve a decisional mechanism. Finally, we verify the validation and efficiency of our model by simulating experiments. To the best of our knowledge, we are proposing a decisional P system for the first time, and the results show that this system is very logical and efficiently processes large-scale decisional tasks. Indeed, the conditional enzymatic numerical P system could achieve prior results more quikly by a factor of 188.28, and a decision tree based on DENPS is the 119.85 times faster than the general serial framework.
GaitMGL: Multi-Scale Temporal Dimension and Global–Local Feature Fusion for Gait Recognition
2024, Electronics (Switzerland)
Blockchain-based privacy-preserving multi-tasks federated learning framework
2024, Connection Science
Chunk2vec: A novel resemblance detection scheme based on Sentence-BERT for post-deduplication delta compression in network transmission
2024, IET Communications

View all citing articles on Scopus

View full text

On the throughput-energy tradeoff for data transmission between cloud and mobile devices

Abstract

Introduction

Section snippets

Related work

Basic throughput-energy tradeoff model

Problem transformation

Algorithm analysis

Performance evaluation

Conclusion

Acknowledgment

Inform. Sci.

Inform. Sci.

J. Netw. Comput. Appl.

Inform. Sci.

Inform. Sci.

Inform. Sci.

Appl. Soft Comput.

Inform. Sci.

Optimal power management of residential customers in the smart grid

IEEE Trans. Parallel Distrib. Syst.

Qos-oriented admission control in hsdpa networks

Netw. Protocols Algor.

A survey of computation offloading for mobile systems

Mob. Netw. Appl.

Mobile data offloading: how much can wifi deliver?

IEEE/ACM Trans. Netw.

Gearing resource-poor mobile devices with powerful clouds: architectures, challenges, and applications

IEEE Wireless Commun.