Measuring eBPF Cache Without Leaving the Kernel
New methods for measuring inode cache statistics of the eBPF agent at the kernel level.
While testing our eBPF agent, I noticed discrepancies between user experiences and my benchmarks, particularly in performance-critical areas. To understand inode caching in user environments, I developed metrics to measure cache usage, including hits and misses. This proactive approach aims to quickly address performance issues by gathering insights directly from the kernel.