I'm new to fio, and its excellent documentation. While I can use it successfully to compare (fio) based loads across various hardware, I'm a bit lost on how to translate an observed production workload, across to fio. My overall objective is to reproduce a production DB load with fio, and see how the busy/latency/queue depth improves for a variety of different improved storage arrangements. What I have in prod is a single network block storage device via atop: - 100% busy 100% of the time - mostly doing 200MiB/s read, 3MiB/s write - avq 66.82 clearly this NBD is insufficient. I've found https://github.com/iovisor/bcc which will be my next step to get more detailed information. Does anybody have other tools, sysctls or metrics to look at, and suggestions on how they can be mapped to fio usage? thanks dch
