Codes with efficient erasure correction

Thumbnail Image


Publication or External Link





Distributed storage systems are becoming increasingly ubiquitous in the emerging era of Internet of Things. Major internet technology companies employ large-scale distributed storage systems to accommodate the massive amounts of data generated and requested by global users. The need of reliable and efficient storage of immense amounts of data calls for new applications and development of classical error-correcting codes.

This dissertation is devoted to a study of codes with efficient erasure correction for distributed storage systems. The efficiency of erasure correction is often assessed by two performance metrics, bandwidth and locality. In this dissertation we address several problems for each of these two metrics. We construct families of codes with optimal communication complexity for erasure correction ("repair bandwidth") for a heterogeneous storage model, and derive several results for the problem of optimal repair of Reed-Solomon codes. We also construct families of cyclic and convolutional codes with locality, extending the range of parameters for which such families were previously known.