I can pretty quickly monitor the running time of a process with time:
x@y ~ $ time foo real 0m14.299s user 0m4.770s sys 0m0.440s Is there a way I can get the same data for I/O and CPU usage of an argument, recorded to STDOUT? A simple command or utility like time would be ideal, where I just pass the argument of the thing I want to run:
x@y ~ $ stats foo wallclock runtime 0m14.299s I/O reads 290,420 KB I/O writes 239,429 KB peak CPU usage 18.62% mean CPU usage 1.44% # etc. look at the time man page on your system, some implementations have format options to include I/O, CPU and Memory stats (-f).
For instance, GNU time, with -v will show all available info (here on Linux):
/usr/bin/time -v ls Command being timed: "ls" User time (seconds): 0.00 System time (seconds): 0.00 Percent of CPU this job got: 0% Elapsed (wall clock) time (h:mm:ss or m:ss): 0:00.00 Average shared text size (kbytes): 0 Average unshared data size (kbytes): 0 Average stack size (kbytes): 0 Average total size (kbytes): 0 Maximum resident set size (kbytes): 3664 Average resident set size (kbytes): 0 Major (requiring I/O) page faults: 0 Minor (reclaiming a frame) page faults: 273 Voluntary context switches: 2 Involuntary context switches: 2 Swaps: 0 File system inputs: 0 File system outputs: 0 Socket messages sent: 0 Socket messages received: 0 Signals delivered: 0 Page size (bytes): 4096 Exit status: 0 On BSDs, use -l instead.
Note that this is the actual /usr/bin/time program, not the keyword that some shells like bash provide which you invoke with time pipeline.
zsh's time keyword can also be configured (with $TIMEFMT) to provide with that information. command time -v ... since bash is hijacking time. The command strace can be useful, you can limit the trace to only count -c or a subset of system calls,
-e trace=set Trace only the specified set of system calls. The -c option is useful for determining which system calls might be useful to trace. For example, trace=open,close,read,write means to only trace those four system calls. Be careful when making inferences about the user/kernel boundary if only a subset of system calls are being monitored. The default is trace=all. In this answer, I use it to count the number of system calls. I don't know if anything will give you a summary breakdown of throughput, however strace can produce the figures for something like awk to summarise.
strace can slow things down significantly, so it's useful to measure relative times of different syscalls, but I would not use it as a general purpose timer. perf is far faster/less intrusive, if you play with sampling frequency options.