Analytics with Intelligent Edge

Wednesday, June 14th, 2017 at 6:30 PM

TI Auditorium

PROGRAM

6:30 - 7:00 PM Networking & Refreshments
7:00 - 8:00 PM Talks
8:00 - 8:30 PM Panel Session
8:30 - 8:45 PM Speaker Appreciation & Adjournment

NOTE: Refreshments FREE for this event!

SPONSOR(S)

Chair: MP Divakar
Organizer: MP Divakar
    

Session Abstract:Continued proliferation of Internet of Things (IoT) and the utility of application-rich cloud computing have pushed computing paradigms to newer ones like edge computing. Processing the data at the edge of the network is not only expedient, efficient, timely and energy-optimal, it is a critical requirement for real time analytics in IoT Networks. Edge computing has the potential to address the concerns of response time and latency, battery life extension, minimum bandwidth utilization, data safety and privacy. Known also as Fog Computing, it has the advantages of edge caching, client-centric control, and agile development to address dynamic needs of the IoT edge.

Comsoc SCV is pleased to present two speakers from Hewlett Packard Labs (HPE) who will address elements of edge/fog-computing hardware design and the software applications that take advantage of this evolving computing paradigm.

Speaker: Sergey Serebryakov

Bio: Sergey is a senior research engineer at Hewlett Packard Labs. His research interests span supervised and reinforcement machine learning. He has expertise in multi-agent systems, natural language processing, time series classification and distributed deep learning. Before joining HP Labs in 2010, Sergey was a post-graduate researcher at Saint-Petersburg Institute of Informatics and Automation of Russian Academy of Sciences.

Title: Deep Learning in an IoT World

Abstract: A wide range of applications from various domains is powered by Deep Learning (DL). A world of Internet of things (IoT) is not an exception – we see high potential of DL in this domain too. But IoT poses unique challenges for DL-powered applications – less powerful compute capabilities, limited memory capacity and network connectivity of edge devices. We will present performance results for a number of DL workloads on a number of systems including those targeted specifically at IoT. We will then discuss approaches to minimize traffic between edge devices and datacenter nodes in an “edge-to-core” setting, when data is collected on the edge and training is performed in a datacenter.

Speaker: Sai Rahul Chalamalasetti

Bio: Sai Rahul Chalamalasetti is a Research Engineer in the Platform Architecture Lab of the Hewlett Packard Labs in Palo Alto, CA. Before that Sai Rahul Chalamalasetti was a Hardware Engineer in the Silicon Design Lab of the HP Server division in Houston, TX. He worked on development of “The Machine” and the next-generation Gen-Z fabric. He is involved in research and development of next-generation server architectures and workload acceleration using GPGPUs and FPGAs. He received the M.S and Ph.D. degree from the University of Massachusetts Lowell in 2009 and 2012, respectively, both in computer engineering.

Title: New Paradigms in Edge Computing for IoT

Abstract: While some IoT applications can be handled via a direct “Things to Cloud” connection, several of them actually present big data challenges given the volume, velocity and variety of inputs being aggregated making sending all the data to the data center implausible.

In addition to data analytics and management, the Things must be controlled. Real time response and control of the Things is one of top reasons to compute at the Edge. How fast do we need to know a child is about to get hit by a driverless car? How soon would we need to know a turbine or pump motor is going to catch on fire? How soon do we need to know a medical instrument is detecting a stroke? This talk will share insights on characteristics needed in IoT Edgeline Systems to enable Deep Compute, Deep Data Capture and Ingest, and Enterprise Class Systems and Device Management