Special Seminar

Topic: 
Conquering MPSoC Complexity with Bio-Inspired Self-Organization
Monday, November 2, 2015 - 4:30pm to 5:30pm
Venue: 
Gates 415
Speaker: 
Andreas Herkersdorf (Integrated Systems TU)
Abstract / Description: 

The complexity of hardware/software systems evolved as THE grand challenge for the effective deployment of multi-/many-core processor platforms in embedded and high performance computing systems. Whether we struggle with MPSoC power, resilience, mixed-criticality guarantees, software debugging or security shortcomings, all these are consequences of the economically driven increase in function integration and growing system complexity. Nature has intriguingly powerful mechanisms to cope with the inherent complexity of living organisms. For example, the visceral nervous and immune systems control vital body functions without conscious involvement. On a higher layer, the somatic nervous system or brain allows cognition, reasoning and learning. Both layers are dominantly founded on the principle of emergent self-organization: system constituents follow few, individual and simple rules which manifest in a desirable overall system behavior (or the opposite, chaotic behavior, which usually is being eliminated in nature through evolution). These principles inspired initiatives like autonomic and organic computing, cognitive sciences and humanoid robotics.

My talk will first provide an overview on the research activities at our institute, which are focused on architectures, accelerators and design methods for application specific MPSoC. I will then show examples for the successful application of self-organization principles on homogeneous MPSoC. Thereby, workload balancing and optimization of power consumption and system resilience were the target objective functions. Evolution, i.e. the identification of efficient local rules and behaviors, is being emulated by means of generic online and hardware-based reinforcement machine learning techniques. I will close with an outlook towards applying these concepts in the contexts of MPSoC diagnosis / software debugging and SDN/NFV (software defined networking/network function virtualization) scaling.

Bio:

Andreas Herkersdorf is a professor in the Department of Electrical and Computer Engineering and also adjunct to the Department of Informatics at Technical University of Munich (TUM), Germany. He received the diploma degree from TUM in 1987 and the Dr. degree from ETH Zurich, Switzerland, in 1991, both in electrical engineering. Between 1988 and 2003, he has been in technical and management positions with the IBM Research Laboratory in Rüschlikon, Switzerland.

Since 2003, Dr. Herkersdorf is director of the Chair for Integrated Systems at TUM. He is a senior member of the IEEE, elected review board member of the DFG (German Research Foundation) and serves as editor for Springer, Elsevier and deGruyter journals for design automation, communications electronics and information technology. His research interests include application-specific multi-processor architectures, IP network processing, Network on Chip, system level SoC modeling and design space exploration methods, and self-adaptive fault-tolerant computing.