A novel server selection approach for mobile cloud streaming service

https://doi.org/10.1016/j.simpat.2014.06.014Get rights and content

Highlights

  • We propose a mobility-aware framework for mobile cloud streaming services.

  • We present Client-driven QoS-oriented Server Selection Scheme (CQS3) to redirects user requests to the “best” server.

  • We propose dynamic server selection algorithms for mobile cloud streaming systems to support user mobility.

  • We design our simulation experiment carefully by applying GT-ITM.

  • The experimental results validate that our strategy adapts well to user mobility with comparable delay performance.

Abstract

Mobile cloud streaming has become very popular recently. However, the centralized datacenter structure may lead to high service delay, especially for real time and high bandwidth streaming services. Deploying cloud edge servers may improve the quality of mobile streaming services in theory, but it cannot well adapt to the user mobility. To deal with this problem, this paper puts forward our MACSS, a mobility-aware framework for mobile cloud streaming services, which provides dynamic and optimized server selection functions to support user mobility comprehensively. Then we further propose CQS3, a novel quality-aware scheme that frequently redirects user requests to the “best” server according to comprehensive service quality grade. Simulation results show that CQS3 scales well and guarantees comparable delay performance. In addition, CQS3 can significantly improve PSNR comparing with CSS and is well adapted to user mobility and channel variation.

Introduction

Recently, mobile cloud computing has received large interests form both industry and academic. It has been built to bridge the gap between increasing complex applications and insufficiency of mobile device capacities. With the development of the mobile devices capabilities, such as CPU power, network connectivity and sensors, mobile devices are becoming increasingly popular, and they are used to get access to Internet applications anywhere, anytime. The Internet report of CNNIC points out that the number of mobile Internet users in China has exceeded 460 million, accounted for 78.5% of Internet end users [1]. Huge numbers of users provide unprecedented opportunities for the development of mobile Internet, thus more mobile applications such as online gaming, navigation, streaming and speech recognition are emerging. As a result, more considerable computational power and energy are required. Unfortunately, due to the restrictions on weight, size, battery life, etc., mobile devices are not competent to execute the complex applications, especially for real time and high bandwidth consumption applications like streaming.

To overcome this problem, mobile cloud computing is introduced to offload these computationally expensive tasks to local cloud, such as Amazon’s EC2 cloud and Apple’s iCloud. Cloud computing provides all kinds of Internet service to end users dynamically with Virtual Machine (VM) technologies [2], [19], which makes the service elastic and extendable easily. Unfortunately, these clouds generally adopt the centralized structure, and their datacenters are far from mobile users, leading to high service delay and low quality of experience(QoE) [3], [4]. To reduce service delay and improve QoE, Satyanarayanan [5] presented cloudlets: trusted, resource rich computers in the near vicinity of the mobile user. Based on this work, Verbelen et al. [6] proposed dynamic cloudlets, where all devices in the LAN network cooperated with each other. Both of these two methods used distributed technology to improve mobile service quality. That is to say, by using the “optimal” server near the mobile user to provide service, network congestion is eased and service response delay is reduced. Owing to the resource restrictions and the mobility, mobile devices cannot provide stable resource to other devices, so the latter cannot yield to the expected results. As to the former, guaranteed quality service can be provided on the condition that the appropriate servers are selected.

In fact, server selection, which redirects user requests to the “best” servers with pre-setting strategies, is one of the key technologies in distributed systems, and many researchers have done in-depth researches of it [7], [8]. However, these solutions work well just in wired networks, not in mobile networks. Firstly, these solutions just help clients to select the “best” server at the initial stage and do not consider server re-selections. Users may have to tolerate the deteriorated QoE caused by their move. Secondly, they are all driven by application servers, not well considering the user perceived quality. Last but not least, they are based on IP address. Nevertheless, this address may not always truly reflect the position of mobile users. One recent study [13] exposed that in cellular networks, IP addresses can be shared across geographical disjoint regions within a short time. Obviously, IP addresses in mobile networks may not contain enough location information, which prevents these solutions from working well. So researchers began to study the server selection scheme in mobile distributed systems, and presented some new solutions [11], [12], [13]. However, the existed works only partially solve the problem caused by user mobility. Thus, it is necessary to propose a server selection strategy for mobile clouds similar to cloudlets.

In this paper, we aim to design a appropriate server selection strategy for mobile clouds to reduce their service delay and improve their quality. To this end, we firstly present a Mobility-Aware Cloud Streaming System (MACSS), a framework to provide high quality services for mobile clouds by selecting appropriate servers dynamically, and the Comprehensive service Quality Grade (CQG), a server selection metric. Then, we further propose a Client-driven QoS-oriented Server Selection Scheme (CQS3), which takes both server-side preference and client-side requirement into account. At last, we design our experiment to verify the scalability and efficiency of the proposed scheme. The results show that, comparing with the Closest Surrogate Scheme (CSS) [14], our strategy significantly improves the Peak Signal-to-Noise Ratio (PSNR), adapting well to user mobility and channel variations while guaranteeing comparable delay performance. The major contributions of this paper are summarized as follows:

  • (1)

    We propose a MACSS framework, which provides dynamic server selection functions to support user mobility and maximize user perceived service quality.

  • (2)

    We present CQS3 for mobile clouds to select the appropriate server near mobile user. Our server selection scheme takes quality of service (QoS) and QoE guarantee into consideration in a poor wireless network environment and makes the best tradeoff between server-side preference and client-side requirement. This makes our server selection scheme suitable for mobile environment.

The remainder of this paper is organized as follows: Section 2 reviews the related works about mobile Cloud and server selection. In Section 3, we propose our MACSS architecture and CQG in detail. Section 4 elaborates the working principle of CQS3. Section 5 describes the construction of a simulation experiment and the experimental result analysis. Section 6 concludes this paper and proposes future research goals.

Section snippets

Related works

This section describes the preliminaries and related works on mobile cloud computing and the server selection for distributed systems respectively.

Mobility aware cloud streaming system architecture

Mobile cloud computing has been introduced to overcome the restrictions of mobile computing, but its centralized structure may result in high response delay. Thus some geo-distributed cloud infrastructures were introduced to scale the cloud service regions, such as nebulas [32] and Federation clouds [9]. Based on these geo-distributed cloud infrastructures, we propose our server selecting architecture named MACSS to improve the user experience. MACSS is an application-layer framework for

CQS3 – a novel server selection approach

CQS3 is driven by the service quality perceived by mobile users, and tries its best to select the edge server that maximizes the experience of mobile users continuously. In CQS3, mobile terminal measures service quality of streaming, and chooses the “best” edge server. The whole selection process can be divided into three stages.

Algorithm 1

The re-selection trigger algorithm

Input:
 The current CQG threshold of application a,CQGathre;
 The detected CQG value of application a served by current edge server i

Performance evaluation

To evaluate the performance of our proposed CQS3 server reselecting strategy theoretically, we carefully designed our experiment, which takes into account of both our mobile cloud streaming and the practical Internet environment.

Conclusion

Mobile cloud computing is introduced to deal with the contradictions between all kinds complex mobile applications requirement and resource restrictions of mobile devices recently. Its central structure may lead high service delay. So distributed mobile clouds such as cloudlets have been presented by deploying servers near to mobile devices to reduce the service delay. But how to select the appropriate server for mobile device is a hard task. This paper aims to deal with this problem. To this

Acknowledgment

This paper is supported by National Key Technology Support Program of China under the agreement of 2012BAH89F03, the Post-70s Scholars Academic Development Program of Wuhan University, the Program for New Century Excellent Talents in University and the Postdoctoral Foundation of China under the agreement of 2012M521478.

References (33)

  • R. Buyya et al.

    Cloud computing and emerging IT platforms: vision, hype, and reality for delivering computing as the 5th utility

    J. Future Gener. Comput. Syst.

    (2009)
  • CNNIC, Statistical Report on Internet Development in China....
  • M. Armbrust et al.

    A view of cloud computing

    Commun. ACM

    (2010)
  • O. Oyman et al.

    Quality of experience for HTTP adaptive streaming services

    IEEE Commun. Mag.

    (2012)
  • H.T. Dinh, C. Lee, D. Niyato, P. Wang, A survey of mobile cloud computing: architecture, applications, and approaches,...
  • M. Satyanarayanan, Mobile computing: the next decade, in: Proc. of the 1st ACM Workshop on Mobile Cloud Computing &...
  • T. Verbelen, P. Simoens, F.D. Turck, B. Dhoedt, Cloudlets: bringing the cloud to the mobile user, in: Proc. MCS ’12:...
  • B. Maggs, Key Algorithms in a Content Delivery System....
  • A. Niyom et al.

    An energy-efficient process clustering assignment algorithm for distributed system

    Simulat. Model. Pract. Theory

    (2014)
  • T. Taleb et al.

    Follow me cloud: interworking federated clouds and distributed mobile networks

    IEEE Netw.

    (2013)
  • P. Wendell, J.W. Jiang, M.J. Freedman, J. Rexford, DONAR: decentralized server selection for cloud services, in: Proc....
  • T. Yoshimura, Y. Yonemoto, T. Ohya, M. Etoh, S. Wee, Mobile streaming media CDN enabled by dynamic SMIL, in: Proc. WWW,...
  • M.M.B. Tariq, R. Jain, T. Kawahara, Mobility aware server selection for mobile streaming multimedia content...
  • Q. Xu, J. Huang, Z. Wang, F. Qian, A. Gerber, Z.M. Mao, Cellular data network infrastructure characterization and...
  • K. Stamos et al.

    CDNsim: a simulation tool for content distribution networks

    ACM Trans. Model. Comput. Simulat.

    (2010)
  • E.W. Zegura, K.L. Calvert, S. Bhattacharjee, How to model an internetwork, in: Proc. INFOCOM, 1996, pp....
  • Cited by (0)

    View full text