Modeling interprocess shared-cache contention on multicore architectures with applications in virtual machine CPU scheduling

Multi-core technology significantly altered the x86 microprocessor architecture with the introduction of shared on-die data and instruction caches. The focus of this research is how shared caches are used and competed for by independent processes. The measurement methods developed use statistical techniques to characterize performance in the presence of extraneous events. A new methodology and technique for modeling the interaction of cores and application processes through shared caching is used. The research presented here uses the competitive Lotka-Volterra equations to model how cores executing different applications affect each other through a shared cache. This dissertation presents the first use of the Lotka-Volterra equations to model parts of a computer microarchitecture. This research bridges two scientific disciplines and opens avenues of research and analytical techniques not previously considered.