Monday, November 18th

1. CmpE Seminar: Array BP-XOR Codes for Hierarchically Distributed Matrix Multiplication by Dr. Şuayb Arslan
   • Start time: 03:00pm, Monday, November 18th
   • End time: 04:00pm, Monday, November 18th
   • Where: AVS Conference Room, BM

   • Abstract: Today, data science led to enormous computation and storage requirements that have transcended all of our expectations. The immediate consequence of this is the distribution of data and associated computation work over large clusters of small processing and storage devices that are typically less capable and slow if not fail-stop. In this talk, a novel fault-tolerant computation technique based on array Belief Propagation (BP)-decodable XOR (BP-XOR) codes is proposed for fault-tolerant matrix-matrix multiplication. The proposed scheme is shown to be configurable and suited for modern hierarchical compute architectures that are equipped with multiple nodes, each having many small independent processing units. The proposed scheme outperforms a few of the well known earlier strategies in terms of total end-to-end execution time in the presence of slow nodes, called stragglers. This performance advantage is owing to the careful design of array codes which distributes the encoding operation over the cluster (slave) nodes at the expense of increased master-slave communication. A fundamental trade-off between end-to-end latency and total communication cost is given. In addition, to be able to address the problem of scaling stragglers, an asymptotic class of array BP-XOR codes based on projection geometry is proposed at the expense of some computation overhead. A thorough latency analysis is conducted for all schemes to demonstrate that the proposed scheme achieves order-optimal computation in both the sublinear as well as the linear regimes in the size of the computed product from an end-to-end delay perspective.

     Short Bio: Suayb S. Arslan received the B.Sc. degree in electrical and electronics engineering from Bogazici University, Istanbul, Turkey, in 2006, with high honors and the M.Sc. and Ph.D. degrees in electrical engineering from the University of California, San Diego, CA, USA, in 2009 and 2012, respectively. In 2009, he was with Mitsubishi Electric Research Laboratory (MERL), Boston, MA, USA, where he conducted research on image processing and machine learning algorithms for biomedical applications. In 2011, he joined Quantum Corporation, Irvine, CA, USA, where he conducted research on advanced data detection and coding algorithms as well as reliability modeling for increased capacity cold data storage and distributed systems. His research interests include digital communication and storage, information, coding and reliability theory, image/video processing, distributed systems, and Internet of things. Dr. Arslan currently serves in IEEE Data Storage Technical Committee, INSIC, SNIA, ComSoc Student Competition Award Committee, and as an Associate Editor for Elsevier Internet of Things journal.

   • View this event in Google Calendar