8 Replies Latest reply: Aug 24, 2014 12:18 PM by Joe Mendola RSS

    EIGRP - Load Distribution

    Nersas Marin - CCIE #48922

      Hello!!! I'm working load balancing traffic with EIGRP between fastethernet and serial interfaces to a destination point. One time that I get the condition and configure variance to 100 (for example) I get traffic sharing to the destinion.
      The problem is that I never get match exact metric (for example 3:1). The closest solution is the one that I attach in the picture:

       

      Ethernet Traffic share count it 41

      Serial Traffic share count it 120

       

      Is it a valid solucion in the CCIE lab?? How much allowance do we have to match the correct answer?? I mean, 44:124 should be a valid solution???

       

      I always try to apply the formula but I never get the exact right number:

       

      3*Metric = Metric

      3 * (10^7/BW + DL1) = (10^7/BW + DL2)

      3 * (10^7/BW + xxx(Whaterver you define)) = (10^7/BW + DL2)

      DL2 = answer

        • 1. Re: EIGRP - Load Distribution
          lp4nb

          What actually is being asked to configure in your lab?

          If they say you have to match the traffic share count as told, then you have to configure in a way to match as it is.

          If you could paste some scenario/configs, that may be helpful.

           

           

          Regards.

          • 2. Re: EIGRP - Load Distribution
            Nersas Marin - CCIE #48922

            It's just a general lab where I am trying to share the EIGRP traffic through 2 links in a ratio 3:1. The modifications have to be done locally in the router (just delay) so that, It's supposed that information is enough. But anyway, I attach you an example where R1 is asked to share EIGRP traffic between f0/0 and ser0/0 in a ratio 3:1

            topology.jpg

            Thanks all for your time and help.

            • 3. Re: EIGRP - Load Distribution
              lp4nb

              as the minimum BW is same, lets change the weights for removing the unnecessary complex calculations.

               

              Now, the default delay at R1 for network 10.1.23.0 is 40,000 U sec from R1-R3-, and 21000 Usec from R1 - R2-

               

              metric from R3 = 3* metric from R2

              256.(40000 + x) = 3.*256*21000

              40,000 + x = 63000

              x= 63000- 40000

              x= 23000

               

               

              so on R1 on interface s0/0

               

              interface s0/0

              delay 4300

              exit

               

              R1(config)#do sh ip route 10.1.23.0

              Routing entry for 10.1.23.0/24

                Known via "eigrp 21", distance 90, metric 537600, type internal

                Redistributing via eigrp 21

                Last update from 10.1.13.3 on Serial0/0, 00:00:09 ago

                Routing Descriptor Blocks:

                  10.1.13.3, from 10.1.13.3, 00:00:09 ago, via Serial0/0

                    Route metric is 1612800, traffic share count is 1

                    Total delay is 63000 microseconds, minimum bandwidth is 1544 Kbit

                    Reliability 255/255, minimum MTU 1500 bytes

                    Loading 1/255, Hops 1

                * 10.1.12.2, from 10.1.12.2, 00:00:09 ago, via FastEthernet0/0

                    Route metric is 537600, traffic share count is 3

                    Total delay is 21000 microseconds, minimum bandwidth is 1544 Kbit

                    Reliability 255/255, minimum MTU 1500 bytes

                    Loading 1/255, Hops 1

               

              Regards

              • 4. Re: EIGRP - Load Distribution
                Nersas Marin - CCIE #48922

                Hello Ip4nb, first fo all, thanks for the time helping me to understand this issue.
                I got the same results as you load balancing to 10.1.23.0, my issue was doing it for 4.4.4.4 (Lo0 interface from R4). I wanted to simulate ratio 3 (fast-ethernet) : 1 (serials) . So far, it should be working in the same way, but I didn't get it. Let me ask you some questions to clarify it.

                 

                When you apply the formula, do you really need to " * 256" ?? If you do it in both sides, isn't it the same???
                In your simulation, you define 40000 for serial connection R1-R3. Is it comming from BW^7 / BW-link Kbps + Delay R1-R3 ???
                In your simulation, you define 21000 for serial connection R1-R2. Is it comming from BW^7 / BW-link Kbps + Delay R1-R2???

                 

                I just want to expand everything in order I can understand the overall.

                 

                Thanks again.

                • 5. Re: EIGRP - Load Distribution
                  cadetalain

                  Hi Nersas,

                   

                  just my two cents: the values  lp4nb used are the delays in tens of microseconds and of course you can let  go of 256 as it is amultiplier on both sides of the equation.

                   

                  Regards.

                   

                  Alain

                  • 6. Re: EIGRP - Load Distribution
                    lp4nb

                    When you apply the formula, do you really need to " * 256" ?? If you do it in both sides, isn't it the same???

                    Choice is yours.

                     

                     

                    In your simulation, you define 40000 for serial connection R1-R3. Is it comming from BW^7 / BW-link Kbps + Delay R1-R3 ???
                    In your simulation, you define 21000 for serial connection R1-R2. Is it comming from BW^7 / BW-link Kbps + Delay R1-R2???

                     

                    yes, the cumulative delay is the delay of serial on R3 s0/1, + delay on serial on R1 s0/0

                     

                    same for the path from R2 to R1.

                     

                    If you expand, a factor of 256*(min BW) is added to both sides of the equation.

                    So, it infact is same thing, as its equal both sides.

                     

                    Regards

                     

                    P.S I didn't know if it was for R4, if its for R4 you can do the same for 4.4.4.4.

                    the catch would be, delay factor for serial s0/0 on R2.

                     

                    metric from R3 = 3* metric from R2

                     

                    (20000 +40000 + x) = 3*(21000 +20000)  [ no need to use, 256, or BW stuff, as they cancel out]

                     

                    60000 + x = 3*41000

                    60000 + x = 123000

                    x= 123000 - 60000

                    x= 63000

                     

                    therfore you can change the delay value accordingly.

                     

                    Message was edited by: lp4nb

                    • 7. Re: EIGRP - Load Distribution
                      Nersas Marin - CCIE #48922

                      Thank Ip4nb for you help, and thanks Cadetalian for your two cents!!!!!

                      • 8. Re: EIGRP - Load Distribution
                        Joe Mendola

                        Hi the result is 23000 but then you put delay 4300.

                        where did you get it from please?