Introducing ebpf_exporter

Cloudflare uses Prometheus to collect operational metrics on hundreds of servers, ingesting millions of metrics per second. They utilize node_exporter for basic system information like CPU usage and memory usage, and cAdvisor for container-level metrics. Histograms are introduced as a superior method for measuring low-level metrics, providing more detailed insights into individual events such as disk IO operations. The use of eBPF is highlighted as a solution to collect these histograms with low overhead, offering safe and efficient kernel execution. Examples of how this technology can be used in practice are provided, including IPC, LLC cache hit rate, run queue latency, and more. Cloudflare's ebpf_exporter tool is open-source and encourages contributions from the community.