[afnog] Load Balancing Two ISP links

[Raymond Macharia]

Hi Mike,
do you have a snap shot of sh int or show controller for your E1 that 
goes down, I think the first step would be to remove that gremlin 
otherwise if it keeps going down often no solution will really be of 
help as it will result in unstable routes.

Regards


Raymond

Mike Barnard wrote:
> Hi Raymond, this is interesting. it does solve the redundancy part, 
> but doesn't address the load-balancing aspect of the two links. the 
> current problem that im facing is that any one of the two E1's 
> unexplainable goes dead and there is a tendency for one pipe to fill 
> up while the other is almost not utilised.
>
> that is why i was looking at something that can handle both redundancy 
> as well as load-balance, using the existing hardware and software.
>
> If all fails, ill play around with this and see how well it will 
> scale. I could try backup load 50 5 at least that ensures that the 
> primary link hits 50%, it brings in the second link and when it falls 
> below 5% it shuts down the secondary link...
>
> Thanks
>
> On 10/15/07, *Raymond Macharia* <raymond at accesskenya.com 
> <mailto:raymond at accesskenya.com>> wrote:
>
>     Hi Mike,
>     I currently use BGP to achieve what you are trying to do however
>     if that seems complicated there is a simple command I used some
>     years back and it worked well for me, below is a template.
>     The lines of interest are the ones that start with the word
>     backup. What the first line did is provide redundancy so when this
>     link went down traffic would be redirected to the other link.
>     The second line is for load sharing and this means that if the
>     first link gets to 80% it invokes the second link and if the link
>     goes down to 60% it shuts down the second link.
>     You can play around with the percentages to get the right mix for
>     you. Note that you still have the two defaults to the two links below.
>
>
>     interface Serial0/0
>      description Internet Link1
>      bandwidth 1024
>     * backup interface Serial0/1
>      backup load 80 60*
>      ip address x.x.x.x 255.255.255.252 <http://255.255.255.252>
>      no ip redirects
>      no ip directed-broadcast
>      no ip proxy-arp
>      ip nat outside
>      encapsulation ppp
>      no cdp enable
>     !
>     interface Serial0/1
>      description Internet Link2
>      bandwidth 512
>      ip address x.x.x.x 255.255.255.252 <http://255.255.255.252>
>      no ip redirects
>      no ip directed-broadcast
>      no ip proxy-arp
>      encapsulation ppp
>      fair-queue 64 256 0
>      no cdp enable
>     !
>     ip route 0.0.0.0 <http://0.0.0.0> 0.0.0.0 <http://0.0.0.0> Serial0/0
>     ip route 0.0.0.0 <http://0.0.0.0> 0.0.0.0 <http://0.0.0.0> Serial0/1
>
>     Mike Barnard wrote:
>>     Hi,
>>      
>>     I have a scenario that i need some assistance and insight as well.
>>      
>>     I have two E1 links, terminating on two separate routers, each
>>     from a different ISP. My current problem is that both links do
>>     have moments of unexplained blackouts.
>>      
>>     I would like to terminate both these links onto one router and do
>>     some load-balancing and fail-over on them. CISCO's OER seems like
>>     the thing that will work this out, though in what i have read,
>>     this needs at least (in my understanding) two routers. The
>>     example given in CISCO's OER manual for deployment on one router
>>     depicts two border routers, one acting as the Master
>>     Controller/Border Router, and the other the Rorder Router
>>      
>>     Has anyone played around with OER and is there a possibility of
>>     getting it to work on one router with two serial interfaces, one
>>     loopback and one ethernet interface?
>>      
>>     Is there any other way to load-balance (and have a failover
>>     provision) on a CISCO 1700 router? I would have done IP CEF
>>     per-packet load balancing but i have no control over the end and
>>     this wont scale well with any link failure.
>>      
>>     Appreciate any assistance.
>>      
>>     Regards
>>
>>     -- 
>>     Mike
>>
>>     Of course, you might discount this possibility, but remember that
>>     one in
>>     a million chances happen 99% of the time.
>>     ------------------------------------------------------------
>>     ------------------------------------------------------------------------
>>
>>     _______________________________________________
>>     afnog mailing list
>>
>>     http://afnog.org/mailman/listinfo/afnog
>
>
>
>
> -- 
> Mike
>
> Of course, you might discount this possibility, but remember that one in
> a million chances happen 99% of the time.
> ------------------------------------------------------------