Paper Title: Understanding Data at Storage Edge: An AI / ML Perspective

Paper Abstract: In contemporary Storage systems, data is often dark and unstructured. It is anticipated that future Storage systems will benefit from computation engines that understand the stored data. In this talk we will demonstrate the feasibility of having an AI / ML engine running at the storage edge. We will show a demo with an in-situ video processing and how availability of the meta-data at the edge can result in a significant reduction of the systems' network traffic. Finally, we will also briefly discuss the potential of utilizing the storage edge processing in a Distributing Learning scheme.

Paper Author: Ned Varnica, Director, Marvell

Author Bio: Ned Varnica received the B.S. degree in Electrical Engineering in 2000 from School of Electrical Engineering, University of Belgrade, Serbia, the M.S. degree in 2001 and Ph.D. in 2005 both from Harvard University, Cambridge, Massachusetts. Since 2005 he has been with Marvell Semiconductor in Santa Clara, California. He held short-term research positions at Maxtor Corporation, Shrewsbury, Massachusetts in 2002 and Lucent Bell Labs, Murray Hill, New Jersey in 2004. He spent the summer of 2003 as a visiting researcher at the University of Hawaii at Manoa, Honolulu. His research interests are in the areas of communication theory, information theory, channel and source coding and their applications to digital data storage and wireless communications. Dr. Varnica received the Best Student of the Class Award from the Department of Communications at the School of Electrical Engineering, University of Belgrade in 2000. He is a co-recipient, with A. Kavcic and X. Ma, of the 2005 IEEE Best Paper Award in Signal Processing and Coding for Data Storage.