LinuxCon + CloudOpen + ContainerCon NA 2015

Systems performance is a primary concern for most people involved in technology. This is particularly true and critical for companies investing significant resources in hardware, expecting Linux to scale well on larger machines, naturally beginning with the kernel. While there are several efforts to automate and integrate benchmarking across a wide range of systems, they are not very well known outside the community. In this talk, Davidlohr will present different alternatives to kernel performance monitoring through automation and will discuss how it can serve to actively detect regressions and new issues throughout the kernel, including how developers can easily test their changes on a wide range of workloads and gather important runtime metrics. In addition, different benchmarks for stressing different subsystems will be compared both at a micro level and for real world workloads.