TODO

- check config parameters

- check if flamegraph is still running before executing entry

- modify memory

- merge_all_flamegraphs

- cli to open data by given time

- save perf script only, for flamescope

- cron to systemd

- how to use time between trace points to trigger perf record

- create a sysinfo log folder

- run-parts

- Makefile, how to delete files

- /proc/PID to trigger perf record or sysinfo

- ceph libunwind libdw

- ceph osd perf counter, use avgtime to trigger log level or perf record

- vmtouch, check some directories

- profile

mem_ftrace)
The next step will need to implement order/migrate-type/size based
flamegraph to visualize the result easily. Also, other report can be
generated:

1). The 5s memory-related data comparison(vmstat/meminfo/zoneinfo).
2). The memory allocation comparison during the 5s with each process
and what type of memory it allocates.