What "time" are you looking to find? Wall-clock time, or CPU time?
None of the netperf tests are set-up to (directly) measure the
unidirectional time it takes to send a packet from netperf to
netserver. That would require smarts that netperf just does not
have. (and perhaps some rather well-synchronized clocks as well...)
One can make a decent first-order approximation of the wall-clock time
by using the netperf TCP_RR test. Take the transactoin rate, invert
it, and then divide by two. This assumes symmetry which may not be
there in anything other than a simple topology, but it would be a
start.
If you meant CPU time, then you want to enable CPU utilizatoin
measurements, and then start doing some arithmatic with the serice
demands.
rick jones