Advanced coding techniques for fast failure recovery in storage systems

点击翻译按钮获取中文摘要

This project aims to improve the performance of distributed data storage systems during the recovery of node-failures using advanced repair techniques for the underlying erasure coding schemes. Reed-Solomon codes, employed in most of current storage systems, for example Google Colossus, Baidu's Atlas, Yahoo Object Store, and Facebook's f4, require extremely high network bandwidth and disk I/O for failure recovery. Expected outcomes of this project include significantly improved repair schemes for Reed-Solomon codes with respect to the network bandwidth and disk I/O. The project will benefit data storage service providers, governments, businesses, hospitals, and schools.