Here are all tools that we can use to classify the traffic under a class-map:
- Existing markings ( CoS,IP Precedence,DSCP)
- QoS group
- Protocol (using NBAR)
- Traffic matching another class-map
- Mac-address (source or destination)
- Range of UDP ports
Just as a reminder , CoS marking (Layer 2) doesn’t “survive” to Layer 3 as the route processor will remark it.
It is why the most of the time a Cos to DSCP rewriting policy must be applied on the input direction of the interface.
Before enabling AutoQoS , take in mind the following pre-requisites :
- CEF must be enabled as AutoQoS relies on NBAR which is dependant of CEF.
- No Policy QoS can be attached to the interface
- The correct bandwidth statement must be set up on the interface
- An IP address must be configured on an interface if its speed is less than 768 Kbps
- The interface can’t be in a shutdown mode
You can the ask why IP Address is an mandatory element ?
It is simply because if your speed links is less than 768 Kbps , then AutoQoS will enable Multilink PPP (MLP)
By default , a class-map will use the match-all criteria so it will tell that you must match all your conditions in your class-map before to classify the traffic .
If you want classify directly when at least one condition is met then use the criteria match-any.
Unclassified traffic will go in the default class.