Research On Fast Data Repair Technology Of Mobile Storage System Based On Erasure Codes

When mobile data traffic is growing rapidly,the cellular system is increasingly difficult to afford massive data requests.The communication mode that is forwarded to the core network through the base station can no longer meet the user requirements of large capacity,low latency and low power consumption in the 5G era.With the improvement of computing and storage capabilities of mobile devices,storing content in mobile devices,and mobile storage systems(MSS)using device-to-device(D2D)technology has become a key opportunity to solve large-scale concurrent requests for massive mobile data.When the mobile device runs out of power or departs the coverage of the base station,the data of popular files cached on the mobile device through the caching technology will be lost,and the user can no longer obtain this part of the file data through the D2 D communication technology.Therefore,the MSS need to repair the invalid data in time,and the redundant fault-tolerant scheme is widely used to maintain the reliability of the system.The traditional replication scheme with the large storage overhead is not suitable for the limited storage space of the MSS.In recently years,erasure codes have become the best solution with its low storage overhead and high reliability.Due to the mobility of devices,the number of mobile devices is dynamically changing at any time.Under the conditions of different size of file and moving speeds,the repair latency corresponding to different erasure cods are different.In addition,the communication range of mobile devices is limited,and D2 D communication between mobile devices is an opportunistic link,which reduces the probability of successful data repair,thereby affecting repair latency performance.Regarding the above problems,this paper focuses on the data repair problem of erasure codes in the MSS.From the perspectives of data repair technology,coding schemes and system parameter selection,it focuses on the impact of device mobility and link on-off conditions on system repair latency performance.Furthermore,the relationship between data repair latency performance and system environment parameters is deeply analyzed,and a relay collaborative data repair plan is proposed to further reduce repair latency.The specific content is developed around the following aspects:Aiming at the matching problem between erasure codes and data repair schemes in the MSS,the repair latency performance of different erasure codes under different data repair schemes is analyzed.First,it elaborates on the characteristics of replication,Maximum Distance Separable(MDS)codes and regenerated codes,and compares the reliability,repair locality,storage overhead,and bandwidth overhead.Then,the repair latency of MDS codes and regeneration codes in instant repair,lazy repair,distributed repair and centralized repair are compared.The simulation proves the optimality of erasure code using lazy repair in the average repair latency.Aiming at the problem of excessive repair latency caused by limited communication range,a lazy repair strategy based on repair interval is designed,and the repair latency under different erasure codes and system parameters is analyzed.Based on reasonable assumptions and modeling of the mobility of the devices,the closed expression of the average repair latency based on the repair interval of MDS codes and regeneration codes are derived.Taking the average repair latency as an indicator,the optimal erasure codes and the influence of the moving speed on the average repair latency under different file sizes is analyzed.The relationship between code length,code rate,repair locality and average repair latency is analyzed by simulation.Aiming at the problems of link on-off and the high repair latency caused by opportunistic links,a relay cooperation scheme is proposed for data repair in the MSS.Uniform distribution is adopted in the selection of the number of relay nodes and data allocation,and the average repair latency expression using erasure codes is derived.And further established an optimization problem for relay node selection and data allocation,and this problem is NP-hard.Subsequently,a greedy algorithm for optimal relay node selection and a heuristic algorithm for the fast data allocation are proposed to reduce the high computational complexity of finding the optimal solution,and the superiority of the relay cooperation scheme in the average repair latency is verified by simulation.