Resource Efficiency and Performance in the Cloud
Kathryn McKinley, Senior Staff Research Scientist, Google
Room: Emerald 1
The promise of cloud computing is affordable, available, and secure computing resources managed by experts. To achieve this promise, providers face end-to-end challenges, including efficiently mapping and scheduling customer virtual machines (VM) on to multicore hardware. Improving, by even 1%, VM performance and packing of VMs on to machines, directly reduces capacity requirements and data center costs. This talk overviews how workload properties inform both mapping workloads and choosing hardware configurations. It describes how engineers use Google’s kernel tracing and visualization tools to optimize performance and dynamically validate scheduling mitigation to Intel’s L1 Terminal Fault security problem.
After technical questions, I will discuss research culture and how myths, such as the sole genius, belie the teamwork that the best science requires. Diverse teams further increase the quality of science in controlled and uncontrolled studies. I challenge you to build diverse research teams and mentor the next generation, so that your team will produce more innovative and impactful work.
Dr. Kathryn S. McKinley is a Senior Staff Research Scientist at Google (2017-present). Her research interests span programming languages, operating systems, cloud systems, and architecture with a focus on performance, parallelism, and memory systems. Her teams’ have produced software systems widely used in industry and academia: the DaCapo Java Benchmarks (31,800+ downloads), the TRIPS Compiler, the Hoard memory manager (used by OS X), the MMTk memory management toolkit, the Immix garbage collector (used by Jikes RVM, Haxe, Rubinius, and Scala), and the SHIM profiler. Her teams’ research has garnered Test-of-Time awards (ASPLOS, OOPSLA (2), SIGMETRICS, ICS), best paper awards (Middlewear, ASPLOS), IEEE MICRO Top Picks awards (4), SIGPLAN Research Highlights, NVMW Memorable Paper, and CACM Research Highlights (2). She has worked tirelessly to improve the success and retention of women and under represented minorities in computing research. She has graduated 22 PhD students and has three sons. Dr. McKinley is an IEEE Fellow and an ACM Fellow.