I am using s_time and s_server to benchmark two computers running OpenWrt
Linux and openssl-0.9.8p. Each host has a Gigabit Ethernet adapter and
they are connected by a crossover cable. I am testing how long it takes
to transfer a 620MB file.
The strange result is a discrepency between using new sessions (-new)
and reused sessions (-reuse). Given the size of the file, I would
expect that its transfer renders any speed up due to reusing session
very small. However, this is not the case.
>From a 30 second test, I get:
650,117,164 bytes read in 44 seconds *without* connection reuse (1
conn)
1,300,234,328 bytes read in 37 seconds *with* connection reuse (2
conns)
Does anyone know what is going on? Another thing I have observed is
that CPU goes down drastically when running the connection reuse case,
after s_time prints "starting." I have looked at the source to s_time
and see that it does a "dummy" download before it starts the clock to
ensure all connections reuse a session. But I don't know why CPU usage
would go down when running the rest of the test (after all, things should
still be crypto-bound).
--
Mike
:wq
______________________________________________________________________
OpenSSL Project http://www.openssl.org
User Support Mailing List [email protected]
Automated List Manager [email protected]
