Efficient Repair Schemes In Cloud Storage Based On Interference Alignment

With the advent of the era of information explosion, variety of applications madehigher and higher demands on the storage and handling ability of massive data.Therefore, cloud storage was proposed as an important extension technology of cloudcomputing. Due to the numerous storage nodes, the situation that storage nodes areunavailable occurs inevitably in cloud storage system. In order to ensure the highreliability of the data, information redundancy and node repair mechanism areintroduced into cloud storage system. It is the focus of the computer communicationsindustry that how to ensure the consumption of system resources as little as possible,while the reliability of data as high as possible during repairing of the failure nodes.Relative to the case of a single node failure has been extensively studied, the case ofmulti-node failure is more common in the actual cloud storage system. Efficientmulti-node repair scheme based on interference alignment in cloud storage is studied inthis thesis. The main research contents and achievements are as follows:Firstly, combined with the characteristics of network coding and multi-node repairmodel, the theoretical lower bound of storage capacity and repair bandwidth areanalyzed and deduced via max-flow min-cut theorem in network information flowdiagram, and then proved its reachability. Meaning that if and only if minimum cut islarge enough in information flow diagram which is correspondence with the problem ofnode repair in cloud storage, a network coding scheme for node failures exists inevitably.Secondly, take advantage of the theoretical lower bound of storage capacity and repairbandwidth, the optimal tradeoff of storage capacity and repair bandwidth curve is drew,and two extreme point of the optimal tradeoff curve are calculated: minimum storageregenerating code and minimum bandwidth regeneration code. Further, a Multi-nodeExact Repair (MER) code scheme which is meet the lower bound of minimum repairbandwidth is proposed by extending the exact Maximum Distance Separable (MDS)code. This coding scheme can repair a plurality of systematic nodes or redundancynodes simultaneously.The repair processes of each failure node interfere with each other when severalnodes are repaired simultaneously. Thus, for2k n,2k1d n1, a multi-nodeexact repair scheme based on interference alignment, Multi-node Exact MinimumStorage Regenerating (MMSR) code, is given. Firstly, according to the general node repair model based on interference alignment, the necessity of interference alignment innode repair problems is illustrated. Secondly, the general construction scheme ofMMSR code is given and the MDS property of MMSR code is proved. Furthermore,synchronous repair and asynchronous repair process is presented. Finally, a datareconstruction program of7,3,5,2-MMSR code is given for verifying the feasibilityand MDS property of MMSR code. The results of simulation and analysis show that theMMSR code reduces the storage capacity and repair bandwidth, while increases the datareliability.