8 Replies Latest reply: Apr 15, 2015 6:35 AM by Mike Gannon RSS

    Regarding EIGRP 'K' Values

    Racharla Chandra Kanth

      Hi All,

       

      Many documents saying that the K-values are as follows:

      • K1 - Bandwidth
      • K2 - Load
      • K3 - Delay
      • K4 - Reliability
      • K5 - MTU

       

      But in reality it doesn't appear to like that. While manipulating the metric formula, i see that Bandwidth and K1 are different values and similarly for all.

       

      While in the eigrp configuration mode we can configure K values to be between 0-255. Now

       

      1.Actually what are these K values & for what purpose these K values are ? Is it just to change the metric value, as per the user needs?

      2.Did anyone had a chance or there was a need to change these K values in real time?

      3.What type of questions can i expect regarding EIGRP metric and K values in the CCNA Exam as well as in a technical Interview for n/w engineer?

      4. And in the wireshark captures i'm seeing one more K i.e. K6, what and where it is used?

       

      K values.JPG

      Thanks in advance.

       

      Regards,

      Chandu

        • 1. Re: Regarding EIGRP 'K' Values
          Dmcneil

          I Think of the K values as "knobs" you can use to help tune the metric calculation of EIGRP to fit your needs. They act as a sort of multiplier for providing preference to specific metric components. For example, maybe your organization values Delay more than bandwidth for a certain type of traffic flow. By modifying the K3 weight you can do this.

           

           

          When you set a K value you are multiplying the metric by that value. So setting "K1 = 1" means "1 * bandwidth". Similarly setting "K1 = 0" means "0 * bandwidth" which effectively removes bandwidth from the metric calculation. A setting in lab environments I've used is to only enable K3 (delay) it results in much easier to manage numbers during the learning phase.

           

          My experience with EIGRP is only through study so I cannot speak to how it is typically implemented in a production network. From browsing the forums here and some books I have come to the conclusion that you rarely have to change those settings.

           

          I have not taken the new exams but typical EIGRP CCNA questions for past exams were similar to "which of the following is used in the EIGRP metric calculation" others may have more insight here.

           

          The K6 metric is used for EIGRP Wide Metrics. You are unlikely to see this even through CCNP material.
          Here is a cisco doc I found online just a few seconds ago: EIGRP Wide Metrics

          • 2. Re: Regarding EIGRP 'K' Values
            Ri0N

            Wonderful explanation by Dmcneil. EIGRP K values are just that, i.e. multipliers to tune the metric calculation and give preference to specific components. Let's look at the (complicated) formula.

             

            Metric = ([K1 * bandwidth + (K2 * bandwidth) / (256 - load) + K3 * delay] * [K5 / (reliability + K4)]) * 256

             

            If you set any of the K values to 0, you can eliminate the component from the equation. Similarly, by incrementing the K value you can increase the preference.

             

            K6 is used with EIGRP wide metrics. As I recall, it is not really used by default but it is included in the formula to provide even more granularity. For comparison, here is the (even more complicated) wide metric formula.

             

            Metric = [(K1*Minimum Throughput + {K2*Minimum Throughput} / 256-Load) + (K3*Total Latency) + (K6*Extended Attributes)]* [K5/(K4 + Reliability)]

             

            P.S. Yes, even Cisco documentation says that K6 was introduced with EIGRP wide metrics for future use.

            • 3. Re: Regarding EIGRP 'K' Values
              Ri0N

              I also don't have real life experience with EIGRP but generally changing the K values is not recommended unless there is a very good reason to do so and an experienced network designer is consulted, as the change can dramatically affect the network. However, as Dante pointed out, only delay (K3) is often used in lab environments.

               

              Knowing the default metric formula and what the K values stand for should suffice. However, I'm not familiar with the new CCNA exam blueprint, so I'm only guessing here.

              • 4. Re: Regarding EIGRP 'K' Values
                Martin

                no need for real work experience, just use what cisco set as default values; that bandwidth and delay.

                • 5. Re: Regarding EIGRP 'K' Values
                  Darren Starr (CCSI, 4xCCNP, 7xCCNA)

                  Here's a screenshot of my slides I've made for my CLN Premium recordings

                   

                  eigrp_dual.PNG

                  • 6. Re: Regarding EIGRP 'K' Values
                    Mike Gannon

                    K values act like multipliers.

                    So in effect if the value = 0 then it has no impact on the metric calculation. 1 affects the metric calculation. 2 doubles the affect and 3 means it has even more impact on metric calculation.

                     

                    you can configure the K values. Leave tos = 1. it was never implemented

                    Router(config-router)# metric weights tos k1 k2 k3 k4 k5

                    Why not configure a lab and change the K values and see what happens.

                    • 7. Re: Regarding EIGRP 'K' Values
                      raymond

                      Just to pipe in here, but cisco actually say not to change the K2, K4 and K5 values in the formula as it can cause issues with network convergence due to the changing reliability and load values on an interface.

                      • 8. Re: Regarding EIGRP 'K' Values
                        Mike Gannon

                        Sure Ray,

                        I agree, but for the purposes of learning in a lab environment it canbe very useful seeing he effects on the metrics, FD, FS etc

                        Mike