Research On Fault-tolerant Technology Of Cloud Storage System In Big Data Environment

In the era of big data,data information is growing explosively,and the users demands for storage resources is growing,but the original capacity of single disk storage lags far behind the requirement for data storage.Therefore,the concept of cloud storage comes into being.It is a part of cloud computing.The emergence of cloud storage has led to a technical change in storage industry,which can satisfy the demand for mass,stable,cheap and safe.is a system,With the support of cluster application,network technology,distributed storage,etc.cloud storage gathers together a variety of different types of storage devices by network and software,and to provide data storage and access service function.However,as the storage size increasing,due to a hardware failure,human mistakes,electricity,fire and other artificial or non-artificial reason,storage node's failure probability is also rising,and the losses of the fault are expensive.Therefore,the system fault-tolerance mechanism is the necessary premise of data reliability and has been extensively noticed by many scholars.Erasure code technology is a relatively reliable method of redundant fault-tolerant,and it can provide a high reliability and high fault-tolerance mechanism of fault-tolerance for cloud storage system.In 2000,Ahlswede et al.put forward the concept of network coding.Dimakis based on network coding and combined with the theory of Max flow-min cut in graph theory to analysis node repair information flow graph,and the theoretical lower bounds of the node repair bandwidth are obtained by induction theory.He analyzed two extremal points on the optimal tradeoff curve,and corresponded to best storage efficiency and the minimum repair bandwidth.He obtained MSR regenerating code(Minimum Storage Regenerating Codes)and MBR regenerating code(Minimum Bandwidth Regenerating Codes),but he have not put forward specific encoding.In this paper,we introduce the existing traditional erasure code technologies in cloud storage system,they are:RAID array code,RS code and array size.Comprehensive analysis of these coding,and combine with storage requirements,discuss the advantages and disadvantages of coding.Based on this we do work as follows:1)Based on the MSR regenerating code,we introduce interference alignment technology and put forward a kind of exact-repair VMSR code.Using the theory to prove VMSR regenerating code is a kind of MDS(Maximum Distance Separable)properties of coding.At the same time,through example of(6,3,5)VMSR to illustrate the feasibility of VMSR code and MDS properties.After comparison the repair bandwidth?the number of helper node of VMSR code and RS code,we obtain VMSR regenerating codes show better performance.2)Based on MBR regenerating code,we propose a clear structure S-M(Symmetric Matrix)MBR regenerating code.It encoded the sourse data by a[Cn2,B]-MDS coding,and then put those data in an symmetric matrix.Choosing d-n-1,S-MMBR regenerating code can achieve the MBR regenerating code repair bandwidth minimum value.In the process of repair,S-MMBR regenerating code only need to transmit data,it does not need any calculation.This S-MMBR regenerating code completed an exact-repair and reduced the computational complexity of the process of failure node reconstruction greatly.Using example to describe the implementation process of S-MMBR regenerating code,and using the theory to prove its feasibility.