[LARTC] Latest techniques for multiple PPP load-balancing

Linux Advanced Routing and Traffic Control

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



> On Sun, Jan 19, 2003 at 01:32:15AM -0000, Andy Coates wrote:
> 
> > I currently have 2 ADSL modems, provided by the same ISP and 
> > terminating on the same gateway (I think having the same gateway is 
> > what is causing the problems, since the nexthop is always 
> the same for 
> > each interface). I've tried using equal cost multipath 
> (using ip route 
> > to set two nexthops with equal weights).  Problem is I believe this 
> > uses route lookups to determine which interface to use, and 
> thus for 
> > sending data to common destinations it only ever uses 1 interface.
> > 
> > Does anyone know if its even possible to send packets over each 
> > interface "equally"?  I read somewhere that you might be 
> able to clear 
> > the route cache after every packet, but I've only seen people talk 
> > about it and don't know if its true.
> > 
> > Can anyone help?
> > 
> > I'm running 2.4.20 on Debian (woody), with both ADSL modems using 
> > PPPoA and with most of the QoS and network queuing modules 
> installed.  
> > AFAIK, my provider doesn't support multilink.
> 
> I've done this in a previous life on a [patched] 2.2 kernel 
> with using ISDN-interfaces, using the 'equalize' parameter in 
> the nexthop statement. Possibly, 2.4 supports this natively - 
> but I cannot test it at home.
> 
> You could try something along the lines of the following 
> though (note the
> equalize!):
> 
> ip route add default scope global equalize nexthop dev ppp0 
> via 12.34.56.78 \
>       nexthop dev ppp1 via 12.34.56.78
> 
> (It won't give you doubled download speed of course, this 
> will only help you while uploading - but only if your 
> provider isn't doing per-connection egress filtering in the 
> upstream gateway)

Hi,

I've tried that as well, but it still only seems to use one interface
when attempting to send data to a specific host (and even trying another
connection to the host results in the packets going over the same
interface).

Seems like its using this route caching to determine which interface to
go over still :(

"ip route" shows:
default equalize
        nexthop via 212.104.130.141  dev ppp0 weight 1
        nexthop via 212.104.130.141  dev ppp1 weight 1

So it at least appears to have set it correctly (it added the weights
itself)

Andy.



[Index of Archives]     [LARTC Home Page]     [Netfilter]     [Netfilter Development]     [Network Development]     [Bugtraq]     [GCC Help]     [Yosemite News]     [Linux Kernel]     [Fedora Users]
  Powered by Linux