تجزیه و تحلیل حساسیت از یک مدل سلسله مراتبی از محاسبات ابری موبایل

Mobile cloud computing is a new paradigm that uses cloud computing resources to overcome the limitations of mobile computing. Due to its complexity, dependability and performance studies of mobile clouds may require composite modeling techniques, using distinct models for each subsystem and combining state-based and non-state-based formalisms. This paper uses hierarchical modeling and four different sensitivity analysis techniques to determine the parameters that cause the greatest impact on the availability of a mobile cloud. The results show that distinct approaches provide similar results regarding the sensitivity ranking, with specific exceptions. A combined evaluation indicates that system availability may be improved effectively by focusing on a reduced set of factors that produce large variation on the measure of interest. The time needed to replace a fully discharged battery in the mobile device is a parameter with high impact on steady-state availability, as well as the coverage factor for the failures of some cloud servers. This paper also shows that a sensitivity analysis through partial derivatives may not capture the real level of impact for some parameters in a discrete domain, such as the number of active servers. The analysis through percentage differences, or the factorial design of experiments, fulfills such a gap.

Mobile computing has undergone unprecedented advances, following a huge growth of smartphones and tablet computers market [1]. In addition, increase in the use of wireless networks (WiFi, 3G, and 4G) caused significant changes on the Internet landscape. In 2012, the percentage of traffic created by mobile devices reached about 10% of the total Internet traffic [2], and a growth of 66% in this traffic is estimated by 2017 [3]. Despite the recent advances, mobile computing suffers from resource scarcity, even on the most modern devices. The most common problems are interruption of wireless connectivity, lack of security, hand-off delay, battery discharge and limited computational power [4]. In this context, a new paradigm named Mobile Cloud Computing (MCC) was introduced recently. This aims to utilize cloud computing resources to overcome the limitations of mobile computing, allowing delivery of more sophisticated and innovative applications to the user. The mobile cloud computing market is expected to reach 45 billion dollars in revenues by 2016 [5]. Considering this financial impact level, it is essential to provide services that can be justifiably trusted, that is, dependable services. Among the dependability attributes of a system, availability is one which relates directly to the proportion of time that the system is found operational. Achieving high availability is important to avoid revenue loss, and other harmful consequences of service outage (e.g., interruption of a remote patient monitoring system). When a company’s workforce needs to move around remote areas to accomplish their duties, the time wasted due to system unavailability may also imply low productivity of its employees. The availability modeling and analysis of mobile clouds require the investigation of a large number of possible events in client, communication, and server domains. This paper presents sensitivity analysis of mobile cloud availability based on hierarchical analytical models and distinct techniques to assess the impact of each input parameter. This analysis aims to identify the bottlenecks for system improvement in a case study. We also use a combined evaluation of results from three techniques which complement each other to deal with the analysis of this system. The results show that the system availability may be improved effectively by focusing on a reduced set of factors which produce large variation on steady-state availability. This paper is structured as follows. Section 2 provides important concepts regarding mobile clouds and also presents a background about sensitivity analysis of stochastic models. Section 3 describes a mobile cloud architecture which is analyzed in this paper, whereas Section 4 presents the availability models developed to represent such a mobile cloud system. Section 5 presents the availability results obtained through analytical models and sensitivity analysis performed through distinct techniques. Finally, Section 6 draws the final remarks.