Erasure Coding in Distributed Storage Systems

Data centers are nowadays highly distributed, possibly over continents, to cope with the huge amounts of data collected these days. To deal with data loss in an environment where hardware failures are very common, researchers have developed new kinds of erasure codes. Erasure coding is a common technique used in all sorts of digital communication, including deep space communication and QR-codes, but has only seen now application in distributed storage systems. Facebook and Microsoft use a new class of erasure codes to build more storage efficient and reliable data centers. In this paper we introduce erasure coding and show how such codes are applied in a distributed storage system.