Eigrp on Wan links

lildeezullildeezul Member Posts: 404
Hello. i am having trouble understanding the percent wieght, and how to effiecently use eigrp over wan links such as frame relay.
I am using Diane and Catherines book, *( BSCI self study)
in figure 3-25 theres is a hub router, and 10 spokes using point to point sub interfaces on each PVC.
THe access link of s0 is 265 kbps, and the administrator bought a CIR of 56 to each PVC or remote location.
It goes on to say that if the Hub router was to commmunicate with all 10 spokes at the same time it would use 560 kpbs, which is over the access-rate of the line.
So the examply has the spoke routers configure there interfaces to 25kbps, becuase there is 10 spokes so 256 ( access-rate / 10 = 25 )

but to my understand i though the bandwidth command only effects the eigrp process, not the actual bandwidth of the line.

so even if the bandwidth is set to 25 kbps on all the spokes router, if the hub router would communicate to all 10 at the same time it will still be 560kbps.
so how does the bandwidth command fix that.

maybe im missing something.
Can someone please help me out on this one.
This Eigrp over wan links is really bugging me out, and my brain keeps getting fried every time i try to make my own understanding..

NHSCA National All-American Wrestler 135lb


  • gojericho0gojericho0 Member Posts: 1,059 ■■■□□□□□□□
    EIGRP will by default use upto half the reference bandwidth if needed. Like you said if you did 10 * 56 you would get a bandwidth configuration of 560. This would be greater than your physical line speed of 256.

    Since, the bandwidth shouldn't be greater than your link speed since you could potentially flood your pipe you would assign the bandwidth of 256 (the actual speed of your pipe) to the serial inteface. Once of the EIGRP bandwidth rules states that both ends of the PVC should have the same bandwidth allocation. To get this you divide your hub end 256 by 10 and get 25.6.

    25.6 ~= 25 and you have symmetric bandwidth allocation between both ends of each circuit without flooding your pipe at the head end
  • gojericho0gojericho0 Member Posts: 1,059 ■■■□□□□□□□
  • kryollakryolla Member Posts: 785
    you are correct in that bandwidth and access rate are different. Bandwidth is also used for QOS. I think EIGRP is a chatty protocol
    Studying for CCIE and drinking Home Brew
  • gojericho0gojericho0 Member Posts: 1,059 ■■■□□□□□□□
    kryolla wrote:
    you are correct in that bandwidth and access rate are different. Bandwidth is also used for QOS. I think EIGRP is a chatty protocol

    It can be chatty, but if implemented properly there's ways minimize the traffic. Cisco even helped out a little bit with additions like goodbye messages from neighbors and improving query responses
  • lildeezullildeezul Member Posts: 404
    Thanks for the info.

    I think i get it...

    for point to point sub interfaces.

    this is my interpretation.
    The CIR of all the pvc's is 560, which exceeds the Access rate speed.
    So We need to configure the bandwidth so that eigrp doesnt exceed the access rate. we also need to take in to consideration that eigrp needs to allocate its bandwidth usage equally. Therefore 256/10 is 25.6 so we can effectly set the bandwidth of all the spokes interfaces to 25 even though its CIR is 56, and the bandwidth of to the hub router to 256.

    But eigrp CAN USE IF NEEDED 50% of the bandwidth configures. so half of 25 is 12.5kbps.
    but 12.5 isnt half of the CIR, so we need to change the percentage of bandwidth EIGRP uses to match half of the CIR.

    56/2 = 28

    bandwidth for each pvc = 25.

    so we need to set eigrp percentage usage of the bandwidth configured to 110%, which makes eigrp use 28kpbs if need. ( which is half of the CIR)

    am i right?

    doesnt sound too confusing now.

    Thanks jericho.. ( i remembered when you helped me out on a CCNA question i had, and now you are helping me out with CCNP.. i really appreciate your help)

    but do you think i got the point to point frame-relay down..

    know i got to read up on multipoint
    NHSCA National All-American Wrestler 135lb
  • kryollakryolla Member Posts: 785
    for point to point sub int there can only be one dlci so that is the bandwidth for multipoint you devide the number of dlci's by the bandwidth and add everything up point to point and point to multipoint and compare to access rate of the physical interface. HTH
    Studying for CCIE and drinking Home Brew
  • gojericho0gojericho0 Member Posts: 1,059 ■■■□□□□□□□
    oops, i apologize. i didn't see pure p2p subinterfaces. My explination was for a multipoint at the hub and p2p at the spokes. So i probably confused you even more :)

    With the pure point-to-point you are correct in using the bandwidth percent command. Its basically the same idea, but you are keeping the PVCs subinterface bandwidth the same as the CIR at the spoke end.

    Chances are at the hub end you are over subscribed meaning the sum of the spoke bandwidth is greater than the physical link speed of the hub (like in your example). You have to adjust the percent-bandwidth accordingly on both the hub and spoke subinterfaces to provide enough bandwidth to allow EIGRP to function, but not take up too much bandwidth that it is causing problems with other types of traffic. In the link I sent above, there is a similar p2p example configuration
  • lildeezullildeezul Member Posts: 404
    yeah i defintely got more confused.

    i am going to keep reading this over and over again, to try to make some sense out of it
    NHSCA National All-American Wrestler 135lb
  • gojericho0gojericho0 Member Posts: 1,059 ■■■□□□□□□□
    haha sorry, with p2p since you can assign bandwidth to each subinterface its good to keep the reference bandwidth the actual CIR of the PVC. That is why they recommend using the percentage to adjust the max amount of pipe eigrp can take

    Typically in most all production environments if its hub and spoke you will see the pure point-to-point methodology.

    BSCI is a tough exam. Take your time and don't rush it. I think the material in this exam the foundation of the CCNP and it is critical to understand it
  • lildeezullildeezul Member Posts: 404
    ok so each pvc bandwidth should be the same as the CIR.

    i am tryin to figure out why the book had 25 as the bandwidth and 56 as the CIR, even though is a P2P frame-relay network

    pg 106 in the authorizes self study guide
    NHSCA National All-American Wrestler 135lb
  • gojericho0gojericho0 Member Posts: 1,059 ■■■□□□□□□□
    ok...in this case we can't have PVC = bandwidth because there are too many PVCs and similar to a multipoint calculation we will need to divide by 10. Does this make sense?
  • lildeezullildeezul Member Posts: 404
    yes it does.

    so if there was like three PVCs then you can set the bandwidth the same as the CIR for that PVC.

    ok yeah..

    this is what the book say.

    - configure each VC as point-to-point, and specify the bandwidth as 1/10 the link capacity becuase they are ten remote sites...

    so thats where they get there 25kpbs from...

    - Increase the eigrp utilization to 50% of actual VC capacity
    56 is the CIR so half is 28.. thats were they configures eigrp to use 110 of the BW..

    ok i seem to get it now.

    So this configuration is the same as the multipoint, becuase it has so much PVC that is actually overflows the access rate-

    Also, i do get how to configure multipoint in which the VC have different CIR.
    you use multipoint subinterfaces for the high capacity CIR (the high ones in common)
    and match the BW with the CIR on the spoke side, and at the hub side, the BW should be the sum of the CIR's..
    and for the low capacity CIR ( the one thats different) you use the point-point subinterfaces, and specify the BW the same as the CIR..

    ex: Router C has int s0 3 PVC with a CIR of 256, and 1 PVC with a CIR of 56.
    i should configure the commands like this.
    Router C      
    interface serial0.1 multipoint
          bandwidth 768
    Router C
    interface s0.2 point-to-point
    bandwidth 56

    NHSCA National All-American Wrestler 135lb
  • gojericho0gojericho0 Member Posts: 1,059 ■■■□□□□□□□
    yay! now see if you can remember tommorow ;)
  • lildeezullildeezul Member Posts: 404

    the thing is, will i be tempted to look above at the previous post icon_lol.gif
    NHSCA National All-American Wrestler 135lb
  • kpjunglekpjungle Member Posts: 426
    lildeezul wrote:

    the thing is, will i be tempted to look above at the previous post icon_lol.gif


    I have alot of conceptual issues with this as well, however i found that the three rules help me alot:

    1.The traffic that EIGRP is allowed to send on a single virtual circuit (VC) cannot exceed the capacity of that VC.
    2.The total EIGRP traffic for all virtual circuits cannot exceed the access line speed of the interface.
    3.The bandwidth allowed for EIGRP on each virtual circuit must be the same in each direction.

    Remember that on a multipoint interface, the bandwidth configured is divided among the PVC's automatically, so if you have a hub (R1) with three spokes (R2, R3 and R4), and your line access speed is 768kbps, the bandwidth on the spokes should be set to 256kbps on R2, R3 and R4, this is to account for rule nr. 3.

    If the line access speed into the cloud is only 512kbps, this should be set (bandwidth 512) on the multipoint interface, and then you set the bandwidth on each spoke to ~170kbps. That way, EIGRP will automatically use 85kbps on each VC.
    Studying for CCNP (All done)
Sign In or Register to comment.