Assessing the impact of Linux networking on CPU consumption

As network interfaces in the data-center get faster and faster, and an increasing portion of services is implemented in software, we wonder how many CPU cycles our servers are dedicating to handling network traffic. In fact, real world measurements always represent the first step to evaluate whether new optimizations are needed, in particular given the claim, coming from some SmartNIC vendors, that this cost can be up to 30% of the total amount of CPU cycles available in a data center. This work describes the design and functionality of a novel tool, Netto, that enables in depth observation and monitoring of the Linux kernel’s networking stack in real time, by exploiting the tracing capabilities of eBPF, an affirmed technology which dramatically enhances Linux’s observability by allowing the dynamic injection of user code into the kernel. It is also shown how a dynamic breakdown of the individual components of the measured networking cost can be built on the fly by collecting and analyzing CPU stack traces.