Understanding Computer Systems Through Architecture-Level Events
Shashidhar Mysore, PhD candidate in the Department of Computer Science at the University of California, Santa Barbara
Wednesday, March 4, 2009
By combining software modules and services developers gain the ability to quickly build very complex systems, but this does come at a cost. Modern multi-language, multi-layer, multi-machine software can be extremely challenging to understand, maintain, debug, optimize, and secure given that no one individual may understand how the whole thing really works. At the architecture-level we have a unique opportunity to observe and study the full behavior of these programs because all software operations eventually must be performed as a series of relatively simple instructions (loads, stores, arithmetic, etc.). However, using these instructions to "see through" a complex computer system requires efficient means by which we can capture the required run-time information, quickly summarize and filter useful information, correlate this information with user-understandable cross-layer behaviors, and visualize the data. Towards these goals, I have developed novel architecture and software methods -- a specialized form of 3D IC technology to extract micro-architectural profile information, novel algorithms and hardware designs to efficiently summarize data streams including program profiles, and data flow tomography techniques to understand the provenance of information and privacy of critical data across different layers of the computer system. In this talk, I will present these approaches and discuss several new avenues for future study.
Shashi Mysore is a PhD candidate in the Department of Computer Science at University of California, Santa Barbara (UCSB). He previously worked as a research assistant at the Indian Institute of Science, Bangalore. His research interests are in computer architecture, specifically in the design of algorithms and hardware accelerators to aid run-time program and system analysis. His work on Range Adaptive Profiling received the Best Paper Award at the Code Generation and Optimization Conference in 2006. In 2007, his work on Introspective 3D Chips was chosen as an IEEE Micro Top Pick and his PhD thesis won the Outstanding Dissertation award at UCSB in 2008. Mysore received his BEng in Computer Science from the BMS College of Engineering, Bangalore. He is a student member of the IEEE.