2009/8/28 Claudio Jeker cje...@diehard.n-r-g.com:
I'm running with this version that fixes the way closed sockets are
handled. Instead of ignoring the last fd it will run handle_connection()
on that fd if the fd is valid and has data. I also fixed a off by one in
the same code.
With this
On Wed, Aug 05, 2009 at 07:17:04PM +0200, Henning Brauer wrote:
* Claudio Jeker cje...@diehard.n-r-g.com [2009-08-05 19:04]:
This needs a bit more work and maybe it would make sense to switch away
from poll to kqueue or libevent. poll() gets inefficient when handling
large ammount of fds.
On Sat, Aug 08, 2009 at 05:08:45PM -0300, Christiano Farina Haesbaert wrote:
I've discovered that my patch reads the wrong tcp table entry, as soon
as I fix this I'll mail the final patch.
I had to add rdomain support to tcpbench so that I can use it in my
test setup. Diff is attached, I hope
I've discovered that my patch reads the wrong tcp table entry, as soon
as I fix this I'll mail the final patch.
--
Christiano Farina HAESBAERT
Do NOT send me html mail.
On Sat, Aug 01, 2009 at 07:33:00PM -0300, Christiano Farina Haesbaert wrote:
Hi there,
Here is a much better diff proposal than my previous one.
- Don't fork use poll instead.
- Collect usefull overall statistics, as % bandwith being used by each
connection, total and average rate since
* Claudio Jeker cje...@diehard.n-r-g.com [2009-08-05 19:04]:
This needs a bit more work and maybe it would make sense to switch away
from poll to kqueue or libevent. poll() gets inefficient when handling
large ammount of fds. But that's maybe for later.
that is DEFINATELY for later if it all.
On Wed, 5 Aug 2009, Henning Brauer wrote:
* Claudio Jeker cje...@diehard.n-r-g.com [2009-08-05 19:04]:
This needs a bit more work and maybe it would make sense to switch away
from poll to kqueue or libevent. poll() gets inefficient when handling
large ammount of fds. But that's maybe for