Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore advanced virtual machine placement strategies specifically designed for modern chiplet-based processor architectures in this 28-minute conference talk from KVM Forum. Learn how current Linux schedulers fall short when dealing with chiplet architectures that distribute CPU cores across multiple chiplets, each containing core complexes (CCX) with shared last-level caches, leading to significant inter-chiplet communication penalties and suboptimal VM performance. Discover a comprehensive enhanced VM scheduling framework that addresses these challenges through three key approaches: lightweight monitoring techniques using hardware performance counters to track cache efficiency, memory access patterns, and inter-chiplet communication metrics; intelligent VM vCPU group placement strategies that optimize initial allocation of vCPU groups and associated contexts like vhost and datapath threads across chiplet boundaries while balancing chiplet locality and minimizing intra-cache contention; and performance-guided assignment of optimal virtual topologies to guests for improved chiplet-locality-aware scheduling decisions. Understand how this framework handles dynamic VM placement behavior in overloaded scenarios and provides practical solutions for maximizing performance in modern chiplet-based computing environments.
