For example, you might have a user who ius running a video-conferencing software app on their desktop, and that software might allow the user to set his own DSCP/IP precedence vlues on the packets generated from that software. The user might think, "Ha, I'm going to set everything to the max so I get full priority". You then have the potential issue of
- voice and video traffic flooding the highest-priority, very low-bandwidth, switch queue which is designed to carry critical network messaging>
- video traffic flooding the priority, low-bandwidth, queue designed to carry voice.
The document you are reading describes how you override those bad values:
1/ identify the traffic type using access lists
2/ use classifiers to match access-list matches to classes
3/ use policy actions to assign values as per the identified classification.
The policy action would identify, as an example,
- voice RTP traffic and assign it ip precedence "6" ("5" in a mixed vendor network), DSCP EF(46)
- video traffic "5" and AF41(34)
- media signalling "3" and AF31(26)
- default traffic "0" and BE(0)