I've compiled netperf on a couple of HPUX 10.20 boxes, and on a Linux
box. I get what look like completely ridiculous numbers for tests to
the Linux box from the HP box on a 100MB/s switched connection, but
realistic numbers from the HP to HP tests. I'll try and summarize
briefly in a table. HP-1 and Linux are connected to a switched network
at 100MB/s, and HP-2 is connected at 10MB/s.
Client Server Connection Measured
(netperf) (netserver) (theoretical) 10^6bits/sec
HP-1 HP-2 10^6bits/sec 6.84
HP-2 HP-1 10^6bits/sec 6.16
HP-1 Linux 100^6bits/sec 0.19 ??????
Linux HP-1 100^6bits/sec no response in >2min.
HP-2 Linux 10^6bits/sec 6.20
Linux HP-2 10^6bits/sec 0.41 ??????
If I try to run tcp_stream_script from Linux to HP-1, HP-1 doesn't
finish any of the tests within the confidence intervals in 60 seconds,
and runs for an excessively long time.
Any ideas?
Michael
------------------------------------------------------------
Dr. Michael J. Rensing | http://wwwhep.phys.uvic.ca/~michael
| mailto:mrensing@uvic.ca
Physics and Astronomy | phone: 250-721-7741
University of Victoria | fax: 250-721-7752
------------------------------------------------------------