EIGRP load balancing

ok, even tho its confusing at first. I need to re read it for about 3x to fully understand it.. anyway..thanks so much.. That cleared some of my questions that I was about to ask in the future, but I think you prolly answered it already.. specially the FS and FD..



and some additional questions about the variance config.

when i type in r #variance ?
it says there that, i have to put an integer, is putting a much higher number increases the variance or should i say the unequal cost path?
[/b]

The over simplified way to look at it is to consider the Feasible Successor rule first (The advertised distance of the feasible successor must be less than the feasible distance of the current succesor route).

Remember that all routes have an AD and an FD. The lowest AD of any route is the successor (best) route. The best route also has an FD, which is higher than it's AD.
For any route to be considered as a back up route and therefore held in the topology table, the back up route's AD must be less than the successor routes FD.
The variance command simply multiplies the successor routes FD by the integer you enter, so (config-router)#variance 2 doubles the current successor routes FD. This means that you may now have some routes who's AD is lower than the FD of the current successor (because you've increased the current successor FD with variance), and voila, you have more than one route to the remote network, even though those routes may not be equal cost

BUT... if you have old / slow EIGRP routers on the network that are still functioning and that also have legacy paths to remote networks, making the variance integer too high means that those old legacy routes might also be placed in to the routing table, which is not desirable. I agree that it's a remote or extreme example, but you do need to try to consider ALL the consequences when making configuration changes.

wow. that was informative.. thanks again.. after reading some topics about eigrp, somehow i finally understand the concept behind it, i just need to practice more..

thanks again.. more power

Here's an example. Note, I'm not going to use real AD or FD numbers otherwise I'd be here 'til Christmas, so I'll use easy to look at numbers so as to make the whole thing easier to digest, board formatting isn't going to help again, I'm afraid, but stick with it:

Router A has 3 possible routes to Router X.

Router A via Router B to router X
Router A via Router C to router X
Router A via Router D to router X

OK, the topology table and the route information:

A > X via B
AD=3000
FD=4000

A > X via C
AD=6000
FD=8000

A > X via D
AD=3500
FD=4500

First off then, best (successor) route = route with the lowest AD.
So A > X via B is best and is placed in the routing table.

Now, back up (feasible successor) route. Refer to the rule in my previous post above...


Do any other routes in there have an AD less than the current successor (best) routes FD? Answer = yes. A > X via D.
So we have a successor in the routing table and a back up route (feasible successor) in the topology table.
The remaining route A > X via C is ignored 'cos it's... well... rubbish!

OK, we have a best route and a back up but they are not load balancing 'cos EIGRP does not load balance across unequal cost paths by default. But we can make it load balance across unequal paths by altering the variance command.
As I mentioned earlier, variance simply multiples the current successor route FD by the number you enter, so in our example, variance 2 would double the successor FD from 4000 to 8000. Now we can use all 3 routes for load balancing, and because variance has been configured away from it's default setting of 1, EIGRP "knows" what our intent is, which is to use all the available paths that now fit the feasible successor rule despite the fact that the costs are not equal.

I managed to add a little colour to the router outputs back on page 1 to help see what's going on, as well as adding some extra explanation text to the post with the Show IP EIGRP Topology output, so you should re-read them again, if you can stand it!!!

Here's an example. Note, I'm not going to use real AD or FD numbers otherwise I'd be here 'til Christmas, so I'll use easy to look at numbers so as to make the whole thing easier to digest, board formatting isn't going to help again, I'm afraid, but stick with it:

Router A has 3 possible routes to Router X.

Router A via Router B to router X
Router A via Router C to router X
Router A via Router D to router X

OK, the topology table and the route information:

A > X via B
AD=3000
FD=4000

A > X via C
AD=6000
FD=8000

A > X via D
AD=3500
FD=4500

First off then, best (successor) route = route with the lowest AD.
So A > X via B is best and is placed in the routing table.

Now, back up (feasible successor) route. Refer to the rule in my previous post above...


Do any other routes in there have an AD less than the current successor (best) routes FD? Answer = yes. A > X via D.
So we have a successor in the routing table and a back up route (feasible successor) in the topology table.
The remaining route A > X via C is ignored 'cos it's... well... rubbish!

OK, we have a best route and a back up but they are not load balancing 'cos EIGRP does not load balance across unequal cost paths by default. But we can make it load balance across unequal paths by altering the variance command.
As I mentioned earlier, variance simply multiples the current successor route FD by the number you enter, so in our example, variance 2 would double the successor FD from 4000 to 8000. Now we can use all 3 routes for load balancing, and because variance has been configured away from it's default setting of 1, EIGRP "knows" what our intent is, which is to use all the available paths that now fit the feasible successor rule despite the fact that the costs are not equal.

ahhhhhhhhhhh... i see.. so by configuring the variance 2, multiplies the Successor so that the other route can fit in to the multiplied AD... now I know....

Ok, the explanation was a kick-ass.. thats what I really wanted to know.. even tho im still confused of some of the concept. But this explanation totally explains it all.. thank you so much!