QOS Traffic

[Paul Smith]

Hi Guys,

I have read the QOS section Odom's 200-125 book. It gave a good overview but I feel like I need a lot more information before I can implement some good QOS.

For example I have been able to mark some https traffic using an access list and class map (10 tcp any any eq 443) and a policy map on ingress of LAN interface. THe policy map marks the same traffic with DSCP AF41. I am seeing the hits on the access list. I set up another policy map on the egress of the WAN interface, same class map but with priority percent 50 command. I can't really tell if my traffic is getting shaped.

Can anyone suggest a good book or article that goes into more detail about QOS?

[Joseph W. Doherty]

IMO, to do "good" QoS, you need to know much and that much often isn't addressed by QoS material. I haven't yet have come across one book (or article), on QoS, that really helps you with QoS beyond something simplistic, such as insuring traffic like VoIP bearer traffic has egress priority over all other traffic. So, unfortunately, I'm unable to recommend anything specific to answer such a broad subject question.

Perhaps, though, other posters might.

[Paul Smith]

Hi Joseph,

Thanks. I was hoping someone of your calibre might contribute to my discussion. I am studying ICND1 but I am already getting obsessed with QOS lol. I am trying not get sidetracked with it but it is so fascinating in terms of what you can achieve with it. I have found this Cisco document which I think is going to help. I am not a big fan of reading PDF's but I can print it out.

https://www.cisco.com/c/en/us/td/docs/solutions/Enterprise/WAN_and_MAN/QoS_SRND/QoS-SRND-Book.pdf

[Joseph W. Doherty · Accepted solution]

Laugh - I prefer reading from paper too.

Anyway, much good information in that documentation. Ditto for other QoS documents that Cisco has published.

That said, what I think is missing is not enough emphasis on the needs of the traffic or the broader concept (at least in my mind) what QoS might encompass.

For example, given a single bulk TCP transfer between two hosts, i.e. no other traffic, can "QoS" be used to "help"? Again (in my mind) I believe things such as addressing TCP's RWIN size and addressing sufficient buffering for TCP's slow start, and issues dealing with MTU, can be used to maximize the flow's transfer rate (and "goodput" rate). I believe you need to understand how "things" interact before you can begin to deal with dealing with concurrent flows, especially of dissimilar types.

Further, with TCP, I believe you need to understand the various "flavors", again before you can "help" them with QoS.

Or, often, I believe, overlooked is the interaction of queuing and variable bit rate traffic. Understanding such is critical to how you manage variable bit rate traffic, especially real-time traffic like VoIP bearer and/or video conferencing.

So, what I'm saying, to really use QoS well, you need to understand network traffic, as that's what you use QoS to manage. I.e. don't limit you QoS studies to just "QoS documentation", also study traffic behaviors.

[Paul Smith]

Thanks Joseph for that insight into your approach. It does illustrate that QOS is not the be all and end all of traffic flow.

The way you broke that down kind of reminds me of another piece of advice someone gave me in relation to 'order of processing'.. The discussion came about when I was trying to get traffic across an IPSEC VPN using an access list while at the same time allow the same subnet to NAT. In the end I had to deny the VPN bound traffic out of the NAT access list.

A different issue but similar in the sense where a broader understanding is required rather than just knowing access lists and tinkering with other individual components.

I guess this is why they teach TCP before they teach QOS :) MTU is also something I would love to become more proficient at, rather than just knowing the ping (no fragment) test.

Thanks so much for your help and passing on your knowledge.

[Joseph W. Doherty]

". . . where a broader understanding is required . . . "exactly!

"I guess this is why they teach TCP . . ." yup, but there too, most teaching of TCP doesn't go into all the different "flavors" nor spend much time on the bug-a-boo of TCP, its flow rate control.

QoS makes sense when you understand the context of the traffic environment.