Workbook 2 - Lab1 - QoS

Just stumbled upon this - in the solution we are told how to calculate the bandwidth needed for 8 calls but we are using the regular Frame-Relay overhead of 4 bytes for this.

In task 1.29 however we are asked to use fragmentation. According to QoS Enterprise Guide on page 1-15 (or 33), it is said that the FRF12 header is 8 bytes long. So going forward, the entire calculation is wrong as it is based on an 4 byte FR Header instead of 8, correct? 

Another thing is when we are talking about the shape values in task 1.28. It's said Cisco recommended value is 95% of the CIR, which is 768000 (* 0,95 = 729600) but the Bc should be CIR/100 and that would be 7680 but in the solution, Mark is using 7600. Just a typo or am i missing something?




  • I would use 8bytes as well for the FR head. I believe it's actually 7 bytes header and 1 byte tail.

    And using 95% of the speed means shaping with cir and mincir (or shape average) of 729600 with be 7296.

  • Ok thanks for confirming -

    on the Shape issue - you wrote Be but im talking about Bc - just a typo on your end? I dont understand how Bc would be anything but CIR/100 according to Ciscos best practice but im happy to learn otherwise since this QoS stuff is hard to get down.

  • Yes bc is correct - 7296. be is 0.

  • Hi Falko

    I felt the same when I went through the solutions. I think it should be 8bytes for FRF.12. Mark also metioned in the same video that the Content manager quoted as: "any values with +/- 10% is acceptable in the exam".



  • Yes but it's 10% in terms of kbps. In terms of percent however you're looking at 1-2% tolerance. And here for exemple if you use 8 byte header you're at about 19% and if you use 4 byte header you're at about 17% and this is a 768K link.

    If we were talking about a 368k link then the percent difference would of course be much higher and I don't think you would fall in accepted values.

  • Regarding of the fragmentation... I can be wrong (and Cisco QoS Guid says I am), but on my understanding we shouldn't add fragmentation to the calculation... Let's say, we have FR 768k. Try this small experiment:

    1. Configure fragmentation only on one side of FR. Lets' say we do packet fragmentataion of 960 bits on the HQ router and don't have it on BR1 router

    2. From router HQ start command: ping Pings have respond because we are sending a very small packets and HQ router doesn't fragment them and doesn't add any "Fragmentation" header. Otherwise, router BR1 would never send respond.

    3. From router HQ start command: ping size 1000. Packets get lost... Because HQ router starts to fragment them and BR1 doesn't understand them as they have "Fragmentation" header and BR1 is not configured for it.

    So, for all packects bigger than 960 router HQ adds header, and for packets less then 960 bits - doesn't...

    How big is voice packet? Does HQ router need to add "Fragmentation" header to them? I don't think so. Do we need to add this header to the calculation? I don't think so.

  • I've been struggling with this too.  Frurstrating thing for me is that I have found 3 or 4 different Cisco docs and they all seem to have differnet numbers they plug in to obtain fhe bandwidth required for a G729 calll over a frame relay circuit.  I I've seen everything from 25.6 to 28.1.  What I have settled on is the TAC Bandwidth calcuator which spits out a value of 26.8Kbps for a G729 call over frame relay.  Hopefully that will keep me within the 10%.

Sign In or Register to comment.