CANCELLED

Code Constructions for Distributed Storage

Iwan Duursma — University of Illinois at Urbana-Champaign

(Joint ADM/RTG Seminar)

Storage solutions for storing data on a single disk or a combination of several disks do not perform well at the scale of data centers that employ thousands of servers. Over the last decade two new families of error-correcting codes have been introduced that better address the requirements for large scale distributed storage of data. We discuss the main parameters for these families and present a variety of algebraic and graph theoretic methods that can be used to construct codes with good trade offs among the main parameters.