Re: [MiNT] Libcurl-ssl / RPM package

[Vincent Rivière <vincent.riviere@freesbee.fr>]

Miro Kropacek wrote:
> > http://www.madboa.com/geek/openssl/#benchmark-speed
> Keith? As you have build those packages, you've got this test build as well?

It seems that the s_time functionality is built into the standard 
openssl executable.

---------------------------------------------------------------------
Here are the results from my RedHat 64-bit server:

$ openssl s_time -connect www.gnu.org:443 -www /
No CIPHER specified
Collecting connection statistics for 30 seconds
tttttttttttttttttttttttttttttttttttttttttttttttttttttttttttttttttttttttt

72 connections in 0.36s; 200.00 connections/user sec, bytes read 30528
72 connections in 31 real seconds, 424 bytes read per connection


Now timing with session id reuse.
starting
rrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrr

65 connections in 0.01s; 6500.00 connections/user sec, bytes read 27560
65 connections in 31 real seconds, 424 bytes read per connection

---------------------------------------------------------------------
Then the results from my PC running Cygwin:

$ openssl s_time -connect www.gnu.org:443 -www /
No CIPHER specified
Collecting connection statistics for 30 seconds
ttttttttttttttttttttttttttttttttttttttttttttttt

47 connections in 0.66s; 71.65 connections/user sec, bytes read 19928
47 connections in 31 real seconds, 424 bytes read per connection


Now timing with session id reuse.
starting
rrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrr

73 connections in 0.05s; 1553.19 connections/user sec, bytes read 30952
73 connections in 31 real seconds, 424 bytes read per connection

---------------------------------------------------------------------

Note that the test is subject to network slowdowns, I believe the speed 
of the progress bar when not stalled is a better indication than the 
total number of connections.

--
Vincent Rivière