LinuxCon + CloudOpen + ContainerCon NA 2015

When simultaneous threads compete for processor shared resources like L3
cache it could result in inefficient space utilization if not properly
managed. In consolidated large scale data centers this affects the key
Quality of Service guarantees. This talk will review various benchmark
data that focus on some of the important performance issues related to
cache contention which until recently Linux kernel has tolerated. With
the Linux Kernel Cache QOS framework based on perf and cgroup, we now
have a lightweight, powerfull tool which gives capability to monitor
cache occupancy and make cache allocations for each thread. The talk
discusses the usage of the interface in webservers, datacenters,HPC, its
limitations and explore usage in virtualization and containers context.
This talk is aimed at anyone interested in monitoring or controlling
processor cache allocations.