Monday, February 28th, 3:00 pm at Jacobs Believed in Me Auditorium

Multi-core technology represents a dominant shift in processor architectures, primarily due to the power and thermal limitations of technology scaling. From an application perspective, we need to leverage parallelism and limit inter-thread synchronization. From the operating system perspective, we need to schedule such applications on available processor cores to meet their timing constraints while maintaining caching performance. In this talk, I will present an operating system framework for scheduling and synchronization, which can efficiently utilize multi-core processors for predictable task execution and provide analytical deadline guarantees. Specifically, this framework will address the issues of effectively allocating fixed-priority tasks to processor cores, and coordinating such allocation with synchronization to reduce blocking penalties.

The proposed solutions have been developed in the context of Distributed Resource Kernels, an infrastructure built as a Linux kernel extension for end-to-end resource management. I will also show the application of these solutions to various domains including automotive systems, avionics controls, autonomous robots, and other cyber-physical systems. Finally, I will discuss open research challenges in this domain and provide potential directions for future work.

Bio:
Karthik Lakshmanan is a Ph.D. candidate in the Department of Electrical and Computer Engineering at Carnegie Mellon University, where he is a member of the Real-Time and Multimedia Systems Laboratory (RTML). His research interests include real-time systems, multi-core computing, cyber-physical systems, operating systems, embedded systems, and wireless sensor networks. His Ph.D. dissertation is on the topic of "Scheduling and Synchronization for Multi-Core Real-Time Systems". He is a recipient of the best student paper award at RTSS 2009, and the best poster presentation award at HiPC 2005. He has authored multiple publications on the topic of multi-core scheduling in real-time systems, and has been a regular external reviewer for real-time and cyber-physical systems conferences including RTSS, RTAS, ICCPS, and ICDCS.