Service Recovery Method For Ubiquitous Device Cooperation

With the rapid development of ubiquitous stub environment and intelligent devices, users can enjoy ubiquitous services with lower price and better quality at anytime and anywhere. Ubiquitous service may be a single service or distributed service provided by several devices.Multi-device cooperation means two or more different devices combine to provide users with various flexible and intelligent services, which is the future trend for ubiquitous stub environment. Generally, the ubiquitous service providing for users includes three aspects: service discovery, service selection and service recovery. Service discovery means to locate the service types provided by the devices. Service selection means to choose the best device set to provide all kinds of ubiquitous services for users. Service recovery means to repair the failed links and maintain the continuity and stability of ubiquitous services.When the service discovery and selection has been finished, users begin to enjoy services. However, some factors will cause ubiquitous service interruptions or jitters. These factors include the variable network topology caused by some of the mobile devices' failing or moving, the decrease of link quality and device performance, the malicious behavior of malicious node. In order to ensure users enjoying the services continuously, we need to quickly recover the failed links and reconstruct the device set. How to select the best recovery service path,decrease the interruption number and provide constant service for users is the main problem of this paper. The research is of great significance.This paper analyzes the current research status of service recovery and discusses the shortcomings. According to different network structures and application scenarios, this paper proposes four kinds of service recovery methods for ubiquitous device cooperation.According to whether the network structure is cluster, this paper divides the research content into two types. The first type is based on the clustering network, the other type is not. The first type includes two methods, one is the service retransmission method based on cluster,which adopts the retransmission mechanism of recovery point. The other is the service recovery method based on trust evaluation, which considers the impact of malicious nodes. The second type also includes two methods, one is the quality of service oriented recovery method, which adopts the local and global service recovery mechanisms. The other is the routing reaggregation method based on the detection of abnormal nodes,which also considers the impact of malicious nodes. According to the trust values, the method excludes the abnormal nodes in advance.The specific studies are as follows:(1) In clustering network, this paper presents a service retransmission method. Firstly, it describes the application scenario and service model and proposes a selection mechanism for recovery point.Multiple recovery points are established in a service path between the service provider and the service requester. When the service transmission is failed, the upstream recovery point rather than the service provider can directly resend the service data to its downstream recovery point.Simulation results show that this method can save the transmission delay,improve the probability of successful transmission and provide users with continuous low cost service.(2) Aiming at a search and rescue scenario in a forest fire based on clustering network, this paper presents a service recovery method based on trust evaluation which adopts Dempster-Shafer (D-S) evidence theory.In order to reduce the impact of malicious nodes or device failure, we need to choose more reliable backup devices to replace the failed devices,recover the interrupted routing quicly and extend the service time. This paper firstly describes the scenario and problem, secondly constructs a double mapping model of service layer and network layer. The service requester calculates the direct trust degree and the recommended trust degree of the backup devices, then uses the evidence combination rule to calculate the comprehensive trust degree. The backup device with the highest trust value will be seclected to recover the service. The simulation results show that this method effectively improves the packet delivery ratio and reduces the service execution time.(3) Aiming at a disaster rescue scenario without clustering network,this paper puts forward a quality of service oriented recovery method.When some of the devices fail and the service interrupts, the method firstly uses the service local recovery. If there are no suitable substitutive devices or the search is beyond a certain range, the method will consider the global recovery. This paper firstly describes the scenario and service model, secondly designs a comprehensive recovery cost function. The function includes needed factors, such as the number of substitutive devices, the delay of recovery service path and the failure ratio. Then the method separately devises the local and global algorithms. The simulation results show that this method can effectively shorten the recovery time,reduce the probability of service interruption and adapt to changing circumstances better.(4) How to avoid or reduce the impact of malicious nodes or device failure, decrease the number of link interruption and make the data transmission more fluent is the object of the fourth research content. This paper discusses the preventive service recovery and presents a routing reaggregation method based on the detection of abnormal nodes. All the devices observe and record the behavior of their neighbors continuously and exchange trust values periodically. The devices with last r ranking will be selected as the abnormal nodes. During routing selection process,each node firstly excludes r abnormal nodes when selects the next hop.Before transmitting the data, the source node will adjust the transmission rate. The corresponding service routing aggregation algorithm is also designed. The simulation results show that the proposed method can reduce the packet loss ratio, especially the number of malicious nodes is large.