On Sun, Feb 26, 2012 at 3:55 PM, Frank Cox <theatre@xxxxxxxxxxxxxxxxxxx> wrote:
>
>> Most people would just look at the router's own bandwidth measurement
>> or the one at the ISP's end if that is available.
>
> Possibly, but that wouldn't break it down by machine. And in that situation
> I'd think a per-machine breakdown would be useful because then you'd know if
> you should be yelling at the kid, the wife or the family dog when you get the
> ten thousand dollar ISP bill. Again, it just seems like the sort of thing that
> folks would want to be able to track in certain situations. But apparently
> not.
Unless you do a lot of local media streaming or network backups, the
per-machine usage should be obvious from the interface traffic. And if
you actually want to control it, you would force everything through a
proxy with user logins - otherwise it is sort of like measuring water
usage by how long everyone is in the shower.
>> I thought what made
>> your case uncommon was that you had multiple machines and multiple
>> routers and wanted the measurements for each pairing even though the
>> packets go over the same interfaces with no inherent separation.
>
> The separation is the gateway assignment or the lack thereof (for local
> traffic). But other than that, yep, that's a correct assessment.
>
>> If you added interfaces and subnets for each route you wanted to measure
>> separately the normal tools would work naturally.
>
> Indeed, but that adds a whole new layer of complexity to my network that's not
> really needed for any other purpose.
A couple of NICs and a switch aren't all that complicated - but the
iptables counters should work. Or you could push one or the other of
your routes though a proxy that keeps its own statistics and factor it
back out of the relevant interface traffic.
--
Les Mikesell
lesmikesell@xxxxxxxxx
_______________________________________________
CentOS mailing list
CentOS@xxxxxxxxxx
http://lists.centos.org/mailman/listinfo/centos
