Application-grained Block I/O Research For Edge Computation

Nowadays,it becomes easily for electric devices to access the network with the development of mobile Internet and 5G network technologies.While the central network,where the great amount of data from the large number of devices gathers,becomes the bottleneck of the entire network.The edge computational architecture which migrates data processing to the edge of the network,is to deal with this problem.Computational task can be migrated at the application granularity with the support of lite-weight virtualization technology.However,we can hardly detect the workload change caused by application-grained task migration in systemic granular.Therefore,in an edge computational system,without the ability to figure out the changes of block level I/O workload,it becomes an unignorant factor limiting the performance optimization.In order to cope with the above problem,we propose the application-grained block I/O analysis method.Moreover,an automatic program framework is developed.The proposed method isolates the I/Os from applications,analyzing individually,thus,the results can be concluded based on above analysis.The contribution can satisfy the demand for block I/O analysis for edge computational.In addition,an application-grained block I/O simulation is proposed for edge computation.The workload is abstracted to three dimensions,a simulation framework is develope,simulation models are integrated and the application-grained block I/O simulator for edge computation is implemented,which provides a reference with a benchmark tool for the study of edge computational storage system.Finally,the evaluation results on application-grained block I/O simulation for edge computation show that the differences between simulated workloads and original workloads are lower than 1%,3%,and 30% individually in the dimensions of global workload stress,time series autocorrelation and sequential access distribution.