Extra Seminar Computer Science - Dr. B. Koldehofe

Title: Distributed Real-Time Analytics: Accelerating performance over Programmable Communication Systems

Abstract:
Real-time analytics is a central building block for distributed applications in the era of the Internet of Things.
Distributed applications can access and develop insights in information stemming from geographically dispersed and heterogeneous data sources, e.g. sensors on mobile devices. This way distributed applications can better understand situations in their environment, and in turn adapt to or even control physical processes. Applications dependent on real-time analytics have challenging performance requirements with respect to the time and quality insights can be provided. Real-time analytics on data streams requires therefore the careful composition of mechanisms and resources in processing and analyzing data streams. At present, most approaches for real-time analytics are oblivious to the cost and overhead imposed on resources and mechanisms of the underlying communication system.

In this talk, we motivate and highlight underlying principles of distributed real-time analytics in the context of event processing systems. We illustrate and pinpoint to performance bottlenecks that require a more flexible use of communication systems. Finally, we present own research showing how programmable communication systems can assist in overcoming performance problems and help in improving the efficiency of distributed real-time analytics.

Short bio:
Boris Koldehofe is senior researcher and lecturer at the multimedia-communications lab, TU Darmstadt. He is currently principal investigator of the DFG Collaborative Research Centre MAKI. In the first funding period of the research center, he also served as managing director and technical coordinator. Formerly Boris Koldehofe worked as senior researcher and lecturer at the IPVS of the University of Stuttgart and postdoctoral researcher at the EPFL. He obtained his Ph.D in 2005 at Chalmers University of Technology. He has extensive research and teaching experience in the area of Networked and Distributed Systems and Algorithms. In particular, he has focused in the past on highly adaptive communication middleware and distributed event-based systems.