Tuesday, November 16, 2010

EIGRP Variance & Unequal Cost Load Balancing Demystified - With Little GNS Bug

It's been a month since I posted my EIGRP challenge lab:


http://deepakarora1984.blogspot.com/2010/10/eigrp-challenge-lab-1.html


So today it's time to show you how to do it. I won't be spending much time today in explaining theory behind all this as so many people have already written about it but of-course that doesn't mean they covered it from all aspects.


But still I'll provide some reference links in the end of this post to make sure you don't have any gaps regarding concepts.


So lets load the initial configurations which I provided in original post and lets get started :


NoteClick Images To Enlarge










so far we can see the route is in routing table of R4, but we don't see redundant path as yet because of it's higher cost.


Lets calculate variance to install redundant route in routing table:


For those who don't understand variance yet here is the link:


http://www.networkworld.com/community/node/29966


Lets first examine the difference between two metrics that belong to both links for same route:




Now to calculate the required variance use the equation: Higher FD / Lower FD, and result will be:




As variance can't be fractional, so we need to take higher value, which will be 4


Lets's use variance value 4 and take a look at routing table:







Looks like both routes now have made it to routing table. So first objective is achieved. Now lets take a look at Traffic Share Count for both links for same route:





so as you can see that current traffic share count is 21:80. Traffic share count is inversely proportional to metric difference between both links.


So next step is to somehow make it as 1:1 :-)


As you might be thinking that Mr. Garcia (Creator Of EIGRP) should be around :-)




so lets take a look at first how much difference is there between both metrics:




Now lets subtract the lower metric from higher metric: By Metric I mean FD AKA RD:




So the difference comes as - 30000128


Now in next step lets add this metric to link with lower metric using "offset-list".








Here offset List calling ACL "0" means matching all routes... just for your sake :-)


Lets check again different EIGRP tables and see if it works:




Duhhh....looks like GNS is creating pain in butt again.... tired of GNS bugs :-(




Actually to over come this issue I had to jump to real equipments with same configuration and here are the results:




Though IOS ver "15.x" was not liking somehow idea of offset-list 0, so I had to create a specific ACL for this route:




So all functional and we have traffic share count of 1:1 between both links though they have different bandwidth.


Some more interesting reads on EIGRP or I should say "Must" read documents on EIGRP Load Balancing:


http://www.cisco.com/en/US/tech/tk365/technologies_tech_note09186a0080094823.shtml

http://www.cisco.com/en/US/tech/tk365/technologies_tech_note09186a0080094820.shtml

http://www.cisco.com/en/US/tech/tk827/tk831/technologies_white_paper09186a00800a62d9.shtml

http://www.cisco.com/en/US/products/hw/modules/ps2033/prod_technical_reference09186a00800afeb7.html/

http://www.cisco.com/en/US/tech/tk365/technologies_tech_note09186a008009437d.shtml

http://www.cisco.com/en/US/tech/tk365/technologies_tech_note09186a00800c2d96.shtml

HTH...
Deepak Arora

No comments: