Coding for Modern Distributed Storage Systems

Lecture 1: Coding for Modern Distributed Storage Systems I
Lecture 2: Coding for Modern Distributed Storage Systems II
Lecture 3: Coding for Modern Distributed Storage Systems III
Lecture 4: Coding for Modern Distributed Storage Systems IV
This series of talks is part of the Information Theory Boot Camp. Videos for each talk will be available through the links above.

Speaker: Alex Dimakis, University of Texas, Austin and Parikshit Gopalan, Microsoft Research

As more and more data moves to the cloud, there is an acute need for efficient, fault-tolerant schemes for data storage. Coding theory offers solutions for fault-tolerant storage that are potentially far more efficient than replication. At the same time, the cloud setting presents some novel challenges that the classical theory does not handle.

There are currently two distinct flavors of erasure coding schemes that address this challenge. Regenerating codes attempt to minimize the total amount of data communicated over the network in order to recreate lost data, whereas locally repairable codes (LRCs) attempt to minimize the number of reads required from other machines. Both lines of work lead to new questions about classical objects such as MDS codes.

This tutorial will be in two parts, each focusing on one of the above two lines of work. The first two lectures, given by Parikshit Gopalan, will be on locally repairable codes; the second two lectures, given by Alex Dimakis, will be on regenerating codes. Both speakers will survey the rapid recent developments in this area and the theoretical and practical challenges that lie ahead.