Re: Call for testers: nsock-engines experimental branch

[Henri Doreau <henri.doreau () greenbone net>]

2011/7/18 Fyodor <fyodor () insecure org>:
> Thanks, Henri!  We've been talking about this idea (using more
> "modern" interfaces in Nsock for a while).  Have you done any
> benchmarking yet?  Have you found cases where the performance of the
> new system is materially better when used by Nmap, Ncat, and/or Nping?
> Cases where it is worse?
Hello,

I am still designing and running benchmarks, and compiling results to
measure performances of this nsock branch. I'll post everything here
as soon as possible.

So far I have already improved a couple things (thanks to the feedback
I received from Luis and long profiling sessions) but it seems that
the new architecture (iterating over the events of each IODs) is not
as efficient as the existing one (traversal of all the events) as we
do traversals of lots of shorter lists instead of a couple big ones.
This is a point I want to improve first, even if the impact is not
that huge (cf. Luis results).

The most obvious advantage I have observed comes from the epoll based
engine which removes the limitation of the maximum number of sockets
nsock can use (up to the system configuration of course).

Stay tuned :-)

> Cheers,
> Fyodor
Regards.

-- 
Henri Doreau |  Greenbone Networks GmbH  |  http://www.greenbone.net
Neuer Graben 17, 49074 Osnabrueck, Germany | AG Osnabrueck, HR B 202460
Executive Directors: Lukas Grunwald, Dr. Jan-Oliver Wagner
_______________________________________________
Sent through the nmap-dev mailing list
http://cgi.insecure.org/mailman/listinfo/nmap-dev
Archived at http://seclists.org/nmap-dev/