I am going to design and implement a performance monitoring solution for my
company's network (LAN/WAN). My requirements include availability and
end-to-end response time. I am looking at netperf to take these
measurements, and it seems OK up to now... but I have however some
questions:
- netperf is clearly designed as a network performance "benchmark" more than
as a continuous monitoring tool. Does anybody actually use netperf to
perform regular tests (from cron for instance) ? Any issues doing this ?
- netperf is quite flexible in terms of protocol selection (I wish to use
UDP), request/response packet sizes, socket buffer sizes and more... BUT is
there a way I could get a round-trip time (instead of the trx/sec) ? Is it
safe to simply invert the trx/sec value for an estimation of the round-trip
time (in sec) ?
- Using the "-P 0" argument, it is quite simple to parse netperf output.
Has anybody done this already ? Wishing to share some code maybe ?
Thanks for such a great program anyway :)
Luc Moreau