Computer Sciences Dept.

Mary K. Vernon

Professor Emerita of Computer Science
Picture of Mary K. Vernon
Quantitative System Design Interests:
Analytic system performance models that accurately predict measured system performance for alternative system design options, important next-generation commercial system design questions, quantitative design of near-optimal hardware and software, near-optimal enterprise storage systems, network traffic characterization, near-optimal network protocols, state-of-the-art parallel/distributed applications and architectures.

Current Professional Activities

  • Consultant, quantitative system design

  • Member, 2020 ACM SIGMETRICS Achievement Award Committee

  • Member, IFIP WG 7.3


Prof. Vernon received the 2019 ACM Sigmetrics Achievement Award in recognition of her record of sustained fundamental contributions to analytic system performance modeling techniques and the use of those techniques to design a wide range of impactful near-optimal hardware, software and communcation system architectures that have known performance properties. Her modeling technique contributions include: (1) customized queueing-theoretic and other analytic techniques that reflect the mechanics of the system and accurately predict measured system performance for various design options, (2) analytic bounds that quantify the opportunity for improvement as well as the target system performance, and (3) customized analytic models that directly derive an optimal system design. She has developed these techniques to address a much broader range of important commercially-relevant system design questions than were previously known to be amenable to analytic modeling, accurately capturing the performance impacts of more complex and detailed system behavior than known to be possible with such abstract models. Moreover, because the simplest analytic model that predicts measured system performance readily yields insight into all performance bottlenecks in a system design, her models have yielded powerful system design insights, including new, and in many cases near-optimal, designs that significantly outperform previous solutions, for: cache coherence protocols, bus arbitration protocols, multi-core/memory interconnection architectures, shared memory architectures for processors with aggressive instruction-level parallelism, system synchronization primitives, operating system semaphore architectures, high-performance parallel job scheduling policies, state-of-the-art high-performance parallel applications - including sparse matrix computations, particle transport codes, large-scale stochastic optimization and real-time ray tracing - running on state-of-the-art large-scale parallel architectures and computational grids, scalable video streaming protocols, distribution networks for popular streaming content, Internet transport protocols, differentiated service Internet link scheduling policies, enterprise storage systems, and flexible manufacturing systems. A key observation from her body of work is that abstract analytic models that predict measured system performance typically require 1/10th - 1/100th, or less, of the development time for accurate system simulators, and a key lesson is that the analytic results frequently uncover significant, previously unknown errors in a complex system (or simuator) implementation. This leads to the important new paradigm that analytic models are also highly valuable for validating simulator and system implementations. Along with her work on analytic modeling and quantitative system design, Prof. Vernon has pioneered the development of new, more precise and insightful characterizations of production high-performance parallel workloads, video server workloads, and network traffic, as well as the development of key new insights into Internet security.

Dr. Vernon's contributions are documented in two U.S. patents for bus arbitration protocols, four U.S. patents for scalable video streaming protocols and content delivery architectures, and over 100 technical papers, including seven award papers - most recently in Sigcomm, Infocom, and the USENIX Security Symposium. Her analytic techniques and design results have been reprinted in selective compilations for system architects, adopted in commercial systems and standards, and used by colleagues in academia as well as at many leading computing companies in the U.S. and overseas. Moreover, hundreds of students she has taught, advised/mentored and collaborated with have gone on to lead or guide major design teams at many of those companies or have joined the faculty at major research universities. Her contributions have also been recognized by a 1985 NSF Presidential Young Investigator Award, a 1995 NSF Faculty Award for Women in Science and Engineering, the ACM Fellow award, the UW-Madison Vilas Associate award, the UW-Madison Kellett Award, and the ACM SIGMETRICS Achievement Award, as well as invited Keynote talks at conferences in the U.S., Germany and Spain, Distinguished Lectures at universities and research institutes in the U.S., Canada and Germany, and numerous further invited talks throughout North America, Europe, Brazil and Japan.

Her further leadership has included serving on the editorial board of the IEEE Transactions on Parallel and Distributed Systems, the interdisciplinary 1993 NSF Blue Ribbon Panel for High Performance Computing, the NSF CISE Advisory Board, the CRA Board of Directors, the interdisciplinary Executive Committee of the National Computational Science Alliance (NCSA), the NRC CSTB Committee on the Internet Under Crisis Conditions: Learning from the Impact of 9/11, numerous program committees for top-tier annual conferences in computer system performance modeling and a wide range of computer, communication, and application design specializations, external review committees for various engineering colleges and computer science departments, and as Program Chair of the 1990 ACM SIGMETRICS Conference, Program Co-Chair of the Performance '96 conference, Chair of the ACM SIGMETRICS Board of Directors, Chair of the Computer Science Advisory Council at Princeton University, and as Chair of the UW-Madison Computer Sciences Department.

Computer Sciences | UW Home