A Novel Method for Combining Algebraic Decoding and Iterative Processing

Abstract

We propose novel error correction coding schemes called generalized integrated interleaving and sparsely integrated interleaving codes. In the context of block interleaved codewords, generalized integrated interleaving allows nonuniform redundancy to be shared among all the interleaves. This allows the redundancy to be adjusted on-the-fly to better suit the error statistics of the channel or storage device. Sparsely integrated interleaving groups data nodes in a distributed storage system into subgroups. A data node can belong to several subgroups. Small errors are corrected locally within each subgroup. A localized algebraic iterative decoding algorithm is used to decode across subgroups to correct large errors that cannot be corrected within subgroups. Very little correction capability is sacrificed to achieve fast error correction and lower communication overhead. This scheme improves data access for all the data nodes and allows easy scaling of the distributed storage network