A Distributed Disaster Tolerance System Based On LDPC Code And Its Performance Analysis

With the sharp increase in the amount of information, enterprises are increasinglydemanding high data availability. When a disaster occurs, how to recover data fully in theshortest time is one of the most concerned problems in the area of industry and academic.Centralized local backup can only deal with not serious problem. But when emergencies likeearthquakes、tsunamis、terrorist attacks occur, the backup data may be not able to be spared aswell as the original data. Therefore, many of important sectors and large enterprises such asgovernment departments、banks、securities centers have built the off-site disaster recoverycenter, that is to make a mirror backup for local data on the remote site. This approach canensure the data availability to some extent, but once both the source data and backup data failssimultaneously, the information data may not be able to recover. Therefore, for some key datathat should not be missing, data availability provided by local data mirroring is notsatisfactory.Erasure code technology is another effective means to improve data reliability in thedisaster tolerance system, which is earlier applied to the local Redundant Array ofInexpensive Disks and communication system. The basic idea is to divide the data into severalblocks, using a coding scheme to generate parity data block. It only need to ensure part of theoriginal data block and parity data block’s availability to recover the original information data.According to the number of data blocks needed to restore the information data, erasure codescan be divided into MDS codes and non-MDS codes. Compared with mirroring scheme,erasure code technology can provide higher availability for enterprise data with relativelysmall space overhead, the cost of which is the time overhead of encoding and decodingoperation.Low-Density Parity Check code was first proposed by Gallager in his doctorial thesis in1963, which is shown to be a good code that can achieve Shannon limit. However, due to itshigh complexity of encoding and decoding plus the hardware condition at that time, LDPCcode could not be widely used.Mackay and Neal’s study in1996catches researcher’sattention to LDPC codes. LDPC codes are now widely used in communications, it is used assatellite digital video broadcast DVB-S2’s universal standard. Due to LDPC code’s non MDScharacter, the paper proposes a fault tolerance performance evaluation mechanism based on minimum erasure list, analyze LDPC code’s fault tolerance performance.Comparing withMinimum Hamming Distance, Minimum Hamming Distance can better reflect non MDScode’s fault tolerance ability. The paper proposes the generating algorithm of Minimumerasure list, and implement by programming in Matlab. Tests the fault tolerance performanceof erasure code of different code length, analyze the effect of code length and check data oncodes’ fault tolerance ability.For parameter of node number、encoding efficiency and decodingefficiency, compare the performance of disaster tolerance model based on erasure mechanismand replication mechanism in theory. The result shows that the data availability of disastertolerance model based on erasure mechanism is higher than the disaster tolerance model basedon replication.Also, the paper presents a disaster recovery scheme based on LDPC code,Comparing with the traditional fully backup scheme and RS coding scheme for large numberof storage nodes, the experimental results prove our scheme can achieve higher dataavailability.