ECE Seminar: New Results on Regenerating Codes for Distributed Data Storage
Monday, November 11, 2013
11:00 AM-12:30 PM
Speaker: Dr. Chao Tian, AT&T Research
Monday, November 11, 2013 at 12:00 p.m.
ECE Conference Room 302, 3rd Floor
Bossone Research Enterprise Center
In large data centers and peer-to-peer data storage systems, traditional erasure codes such as Reed-Solomon codes incur a high repair cost when a node fails, and thus new repair-efficient codes are urgently needed. Despite extensive recent efforts, it remains an open problem whether the cut-set outer bound can indeed be achieved. In the first part of the talk I provide an answer (in the negative) to this problem by completely characterizing the rate-region of the code under a specific set of parameters. The main difficulty in establishing this result is the converse part, and the traditional manual proof approach used by information theorists for the past 60 years appears quite infeasible here. A computer-aided proof (CAP) approach is thus developed, by taking into account the symmetry in the problem, and moreover utilizing the linear programming duality, to yield an explicit algebraic proof. In the second part of the talk, I briefly discuss a new code construction based on layered error correction and combinatorial block designs, which can achieve new operating points that are not possible using existing constructions.
Dr. Tian received the B.E. degree in Electronic Engineering from Tsinghua University, Beijing, China, in 2000 and the M.S. and Ph. D. degrees in Electrical and Computer Engineering from Cornell University, Ithaca, NY in 2003 and 2005, respectively. He was a postdoctoral researcher at Swiss Federal Institute of Technology at Lausanne (EPFL) from 2005 to 2007, and then joined AT&T Labs-Research, Florham Park, New Jersey. Dr. Tian received the Liu-Memorial Award at Cornell University in 2004, and AT&T Key Contributor Award in 2010 and 2011. He is an associate editor of the IEEE Signal Processing Letters, and an adjunct associate professor at Columbia University.
John Walsh, Ph.D.