| With the rapid development of information and communications technology,the global data has exploded.There is a higher requirement for the data storage capabilities of networked systems.Compared with traditional centralized storage systems,the distributed storage systems have the advantages of low cost,large storage capacity,storage system development and fast speed of parallel processing.And the distributed storage systems are gradually received wide attention by the academic and industrial.In order to ensure the reliability and stability of the whole system,distributed storage systems generally use redundant storage strategies such as replication or erasure codes for the data files.However,there are also some problems such as storage resource waste and excessive bandwidth overhead,which greatly reduce the efficiency of system.In recent years,there is a new network coding scheme,the regenerating codes.Since the regenerating codes can significantly reduce the storage cost and bandwidth cost,it is gradually being paid attention to in the field of distributed storage research.Regenerating codes is considered as an effective solution to improve the efficiency of distributed storage systems.However,the regenerating codes is still in the theoretical research stage,and there are two challenges in the application of distributed storage system.Firstly,the computational complexity of regenerating codes is so high,which will lead to reduce the efficiency of data download and repair.Secondly,due to the constraint conditions of network resources,the optimal parameters cannot be selected theoretically by regeneration code.The inappropriate parameters will lead to waste resources,and they will also lead the failure of loading and repairing data.This article mainly focuses on the two aspects mentioned above,the research content includes:Firstly,the storage and flow of data in the system were analyzed through the information flow graph model.The eclectic relationship between storage overhead and bandwidth overhead was given.And got the lower bound of storage-bandwidth overhead.According to the lower bound curve,the minimum storage cost and minimum bandwidth overhead were obtained respectively.Then the construction of the regenerating codes was computed based on the product matrix.Compared with traditional scheme of replication and erasure codes,verified that the regenerating code has the best performance on storage-bandwidth.It laid the foundation for the study of optimization scheme of the complexity and applicability of the regeneration code.Secondly,aiming at the problem of high complexity of regenerating codes,analyzed that the operation of encoding,decoding and repairing of the regeneration code.And then deduced the expression of exclusive-OR operation times in each link,it was found that the amount of computation concentrated in finite field.Through using BASIC(Binary Addition and Shift Implementable Convolutional)instead of traditional finite field operations,the computation complexity of regeneration code was reduced effectively.And compared with the traditional coding scheme,this paper further demonstrated the superiority of XOR and displacement operation for reducing complexity.Thirdly,in view of regeneration code selection within the limited resources,put forward two indexs of unit storage overhead and unit bandwidth overhead.And these indexs provided the theoretical support for selecting appropriate regeneration code under the limited resources.On this basis,this paper proposed a new shortened regenerative code coding scheme based on BASIC.And this scheme can effectively reduce the unit bandwidth overhead.This scheme adapts to the network with limited bandwidth.In this way,the adaptability of regenerating codes is expanded. |
PDF Full Text Request