Holistic resource allocation and analysis for multicore virtualization

This paper presents a holistic framework called vC2M for CPU, cache, and memory bandwidth allocation and analysis in multicore virtualization real-time systems. vC2M integrates shared cache allocation with memory bandwidth regulation mechanisms to mitigate potential interference among concurrent tasks, thus providing better worst-case execution times and timing isolation among tasks and virtual machines. Building on this capability, vC2M further provides a novel resource allocation algorithm for tasks (VMs) and associated theoretical analysis that considers the complex interactions between all three types of resources—including CPU, cache and memory bandwidth—as well as their impact on schedulability. Our evaluation on our prototype shows that vC2M can be implemented with minimal overhead, and that it provides substantial benefits in reducing tasks’ worst-case execution times. In addition, the results show that our analysis is highly efficient, and that our resource allocation algorithm can offer close-tooptimal schedulability performance and outperform a baseline algorithm in both performance and efficiency metrics.