lmbench's bw_tcp used to get fairly high numbers (40? MB/sec)
and now is getting figures ranging from 0.38MB/sec to around 1 MB/sec.
also, a strace of lmbench shows a LOT of
setsockopt(6, SOL_SOCKET, SO_RCVBUF, [114688], 4) = -1 EINVAL (Invalid argument)
and similar, and
the first write(4, "3145728\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 1048576) = 1048576
takes noticably (very noticably, just by watching the strace)
longer than the others.
Simon Karpen slk@shodor.org
Sysadmin, Shodor Education Foundation
"On two occasions I have been asked [by members of Parliament!], `Pray,
Mr. Babbage, if you put into the machine wrong figures, will the right
answers come out?' I am not able rightly to apprehend the kind of
confusion of ideas that could provoke such a question."
-- Charles Babbage