Research On The Collaboration Method Of Mobile Cloud Computing Under Harsh Environment

With the widespread use of smart phone under harsh environment(e.g.,military operation and disaster relief),dealing with the limited capacity of mobile nodes is the key problem.Mobile cloud computing(MCC)has become an emerging computing paradigm to solve the problem of limited capacity of mobile devices by leveraging the “inexhaustible” cloud resources.The key idea of MCC is offloading the overloaded tasks of computation and storage in mobile device onto remote cloud datacenter so as to improve processing speed and save energy.In this paradigm,one of the core issues is conducting collaboration between mobile side and cloud side,which is able to achieve full use of resources and improve the effectiveness of task processing.However,the traditional research is mainly for urban application scenarios,assuming that the network infrastructure is perfect,the remote cloud resources are readily available and reliable.This assumption is no longer applicable to the applications on tactical edges and other harsh environments.In addition,most of the existing research in collaboration only concentrated on task offloading problems.To this end,taking the tactical edge as an example application scenario,this dissertation studies the problems of cooperative mechanism of mobile cloud computing at different levels and different service types under harsh environment so as to deal with the problem faced.The main contents and contributions of this dissertation are summarized as follows.Firstly,a three tiers(remote,proximity and local)mobile cloud collaboration framework is proposed to fit the characteristic of harsh environment.Traditional research is geared towards urban applications,primarily by offloading workload from local device to remote clouds to improve the capabilities of mobile devices.However,for the harsh environment like tactical edge,due to the network dynamics and environmental hostility,remote cloud is no longer available all the time.To solve this problem,based on the specific requirements for mobile systems in harsh environment,this paper designs a three tiers mobile cloud collaboration framework consisted of remote data center,proximity cloud and local mobile device to cope with the harsh environment.It aims to make use of all available resources for task processing,data transmission and storage.The cooperation model of each tier is expounded and the mechanism of dynamic workflow reconstruction is proposed to meet the requirements of fast and dynamic reorganization of cooperative relationship,which lays the foundation for mobile cloud computing under harsh environment.Secondly,a fault-tolerant collaborative mechanism for task processing within the mobile device cloud is proposed.The collaboration in mobile device cloud is an effective way to solve the bottleneck of a single mobile device when remote cloud and near cloudlet are not readily available.Although few studies have considered the instability factors of network,they did not consider the unreliability of mobile node which may lead to the failure of tasks.To address this issue,this dissertation incorporates the reliability of task execution into the modeling of collaboration offloading problem,with the goal of improving the success probability of task execution at fragile mobile nodes.In addition,existing research often assumes the amount of data transmitted per contact is fixed,which does not hold under the situation with dynamic network.In this dissertation,we model the data transmission behavior with multiple contacts,which supports multiple contacts to completely transmit the data.Based on the model above,we propose a dynamic decision algorithm on data and task collaboration offloading based on probabilistic path and a dynamic task replication that dynamically decide the replica number of each task according to the right state.Experiment results show that the proposed algorithm increases the probability of success execution about 30% and reduce the energy consumption about20% comparing with some typical approaches.Thirdly,a multi-cloudlets collaborative mechanism optimization for data fusion tasks is proposed.Multi-cloudlet collaboration is an effective way to solve the problem of burst tasks with dependency at mobile end.It improves the performance of tasks processing in terms of speed and cost by leveraging resources from multiple cloudlets.We take the data fusion application at tactical edge as an example and design a multiple-cloudlets collaboration mechanism based on MapReduce framework.This dissertation models the problem as a collaborative decision making including task allocation,resource provisioning and reducer selection with consideration of cost of processing,transmission,storage and latency.On the basis of Lyapunov technique,an efficient online algorithm is designed.Experiment results show that the proposed algorithm is able to adapt to the dynamic demand for data processing and significantly outperforms some existing algorithms in terms of cost and latency.Fourthly,we design a collaboration mechanism for dense data sharing within mobile device cloud.Data sharing is one of the main operation under harsh environment.Considering the fact that the network under harsh environment is with limited bandwidth and unstable state,a D2D(Device-to-Device)cooperative data sharing mechanism based on mobile device cloud is proposed to reduce the pressure of the central node as well as improve the efficiency of data sharing.The problem of collaborative data transmission among mobile nodes is modeled as a long-term utility maximization optimization,with the constraints of cooperation incentive mechanism and energy consumption consideration.In addition,considering that different mobile nodes have different roles in cooperative group,the timeliness of data acquisition often depends on its roles and appears to be heterogeneous.Therefore,we incorporate the QoS heterogeneity into the design of the cooperative decision model.With pruning the complexity of original problem and exploiting Lyapunov optimization technique,an efficient collaborative strategy that makes both of data transmission and interface selection is designed.Experiments show that the proposed algorithm is able to allocate the network resource to the node with higher QoS in priority and outperforms the existing algorithms in terms of system utility and energy consumption.