I watched the extended webinar, and thank you so much. I was immediately able to take this knowledge and apply it in the real world. I had to build a QoS policy in windows with a decimal value of 26 for outgoing Cisco Jabber packets that were being reduced to CS0 and causing quality issues. I'm still in the process of resolving the issue overall, but having this solid understanding of DSCP has helped me make huge strides just today, and identify the issue and figure out how to correct it. Once again, Thanks, Kevin.
Great tutorial, I just finished my exam in QoS and got 76 points out of 80 points! But you can always learn something every day! Im not 100% sure I agree with you that AF23 is gonna be dropped before AF11. AF23 is still a higher major class than AF11. The minor class (drop pref) is the deciding factor in each specific major class. That means that AF23 should be dropped before AF22 and AF21 but not before AF11. AF23 being dropped before AF11 makes no sense at all. Did you make a mistake Kevin or am I skating on thin ice? I want to get some clarification on this.
Thanks for the quick response! Yeah but thats the whole point of WRED isnt it? So when you configure WRED you dont want the AF23 to be dropped before AF11 but before AF22 and AF21. I see it as an enhancement to instead of treating similar AF1 traffic the same (as IPP) you can use AF11, AF12 and AF13 for "less important" traffic in the same class. Lets say you configure HTTP as AF11, FTP as AF12 and telnet as AF23 and SSH as AF21. Example: HTTP and FTP both AF1 traffic. But you want HTTP to be slightly more important so you put it as AF11 and FTP as AF12. But you still want lets say telnet as AF23 to be more important than all of the AF1Xs but less important than SSH at AF21. Should congestion occur BETWEEN classes, the traffic in the higher class is given priority. If congestion occurs WITHIN a class the packets with the higher drop precedence are discarded first. The only time AF23 is getting dropped before AF11 is maybe if both of them are in the same class which by me is a weird configuration. Thanks for taking your time.
Yes I can agree that WRED does not concider the 3 left-most bits of the DSCP field and that it is the default. But Im just sayin that the whole point of having the drop bits is to differentiate the traffic WITHIN a class. You will never find a provider treating your AF3 the same as AF1 for example. So the default behaviour is kinda useless. Maybe you could just clarify next time that this would never happen in a live production network and give an example of how it might work in production networks. It might get confusing if you dont know how you use it in different classes and how you can tune it. Great video nonetheless. Thanks for clarifying. Good to talk to you, have a nice day.
Exactly! I agree with everything you just said. That was my whole point. I just think we are seeing it from different points of view. Im not disagreeing with you about how it works. I just want to clarify that usually traffic (example AF23) wont be in the same class as (example AF11) hence AF23 wont be dropped before AF11. But yeah, if they are in the same class it works exactly like you explained. English is my second language so I might not express myself exactly the way I want. No hard feelings ofc lol. You can always learn something new especially about QoS. Thank you, Im actually done with my studies now. I got my CCNP in Route, Switch and Tshoot.
Hi, Mr.Kevin. Your videos on CCNA and CCNP level are awesome.They way you present the information are much better than than any other instructors. Currently i am seeing your STP webinar. One thing that confuses me is "What is Network Segment?" This confuses me a lot because many other authors even do not use this term and their diagrams on STP is quite different. Kindly explain in detail so that no confusion remains. I know this is not the comment for this video but i want to ask. Thanks.
Let's say i want to use qos for gaming... should i use the value 46 as well?, since i don't really know what would be a specific dscp for gaming. Grea video by the way :)
Is there any advantage of using L2 marking compared to L3 marking? Or I guess the question should be is there any advantage of configuring QoS based on L2 rather than L3?
In that case, why would anybody choose to do their QoS configuration based on L2 marking. Why not use L3 marking right from the start even within a VLAN, and then you don't need to worry about remarking the L2, if the packet would leave a VLAN. Is there a case where L3 marking could not be used, and you would be forced to use L2 marking?
It is best practice to do QOS over layer 2 rather than layer 3. The reason for this being if you have every edge switch doing the marking, it distributes the load and allows you to delegate your resources more efficiently. If you were doing the marking via layer 3, that means you are more than likely having 1 or 2 routers doing the marking. Its a comparison of having lets say 20+ devices using their resources for marking, to 1 or 2 devices doing the marking and needing to use a lot of those layer 3 devices resources. Make sense? Its about spreading the load.
