[ previous ] [ next ] [ threads ]
 
 From:  "Chris Buechler" <cbuechler at gmail dot com>
 Cc:  m0n0wall at lists dot m0n0 dot ch
 Subject:  Re: [m0n0wall] M0n0 on Alix: Performance Problems
 Date:  Sat, 9 Feb 2008 15:17:27 -0500
On Feb 9, 2008 5:34 AM,  <lutz dot niederer at gmx dot net> wrote:
> Hello,
>
> I just migrated from 1.233 on Wrap to 1.3b9 on Alix.  It was incredibly smooth!  Set up the new

klasse Sache!
>
> Does the other way aroung work, too?  That means, save the config on Alix and push it onto the
Wrap.
>

Yep, you just have to replace the interface names (vr0 -> sis0, vr1 ->
sis1, etc.).


> And I have a weird problem.  The new Alix should now be really fast compared to the Wrap.  But I
do not get any further than about 6,000kb/s.  The line has a max of 20,000kb/s and when I put the
Wrap back on line I get about 17,000kb/s.  I think something must be wrong but I have no idea,
especially as both configs are the same.  Can anybody help with this?
>

Any errors on Status->Interfaces?

What type of WAN connection?


> And while I'm writing this the Alix hung, that means the Alix worked but no packets went in or out
anymore.  I had to reboot.  How stable is the 1.3b9 release on the Alix board?
>

Haven't seen any issues on 4 I now have in production on 1.3b9 in a
variety of environments.


> And a last one: What is the better way to go, use interrupts or device polling on Wrap and Alix?
>

Depends. Polling is slower, but it also ensures the interrupt load
will never make the webGUI inaccessible. Under very high load without
polling, the webGUI won't work. With an ALIX it'll be hard to get it
up to that kind of load (talking 80-90+ Mb likely, I haven't yet
tested to see exactly).

I run all my ALIX boards without polling, but if they were in
environments where they could see 80+ Mb loads I would consider
polling.

-Chris