Bigdata Tolerance Optimization On Cloud Storage Systems

Abstract

Required balance between performance and fault tolerance can be determined by the users of cloud storage
that have usually assign with different redundancy configurations (i.e (k,m,w)) of erasure codes. Here our study finds
that with very low probability, one scheme of coding that can be chosen by thumb rules, for a given redundancy a
configuration which performs best. In this project we introduce CaCo, known as an efficient Cauchy coding approach
for storing information in cloud systems. Initially CaCo makes use of Cauchy matrix heuristics to generate a matrix
set. Later for each matrix in the produced set, CaCo seeks XOR schedule heuristics to produce series of schedules.
Lastly, CaCo chooses the shortest one from all the generated schedules .in this way for an arbitrary given redundancy
configurations CaCo has capability to identify an optimal coding scheme, within the ability of present state of art .by
taking the advantage of caco such as easy to parallelize we can significantly increase the performance through the
selection process with enormous computational resources in the cloud based systems. We incorporate CaCo in
Hadoop Distributed File System (HDFS) and estimate its performance by doing comparison with “Hadoop-EC”
developed by Microsoft research. Our experimental analysis illustrates that CaCo can possesses an optimal coding
scheme within worthy acceptable time. In addition CaCo exceeds Hadoop-EC by 26.68-40.18% in the encoding time
and by 38.4-52.83% in the decoding time at the same instant.