Switch Config best practices

[RobB_8]

I have a question about the best practices for several Switch options... but generally, I assume it is ideal to use the same options on all switches ie:
1) QOS- if you have a statment such as qos dscp 101110 for a high priority vlan, I would assume for the traffic to be properly queued on all the switches, the statement should be on all the swiches that have that vlan, not just on a core switch
2) dhcp-relay and ip helper- I assume if you are using a central (off subnet) dhcp server it is ideal to ensure both statements exist on all switches, not just a core switch
3) ip route- I assume if you are using routing on a core switch, it is the only one that requires the ip route statements, and for performance the fewer the number of routes the better the switch performance (particularly on lower end switches ie 2600's)
Are all these assumptions correct?
Thanks

[Dmitry G. Spitsyn]

Hi RobB !


1. The command:

qos dscp-map < codepoint > priority < 0 - 7 > [name < ascii-string >]

configures an 802.1p priority for the specified codepoint
and, optionally, an identifying (policy) name.
In your case the specified DSCP Policy 101110
implies 802.1p Priority 7 by default.
It means that IP packet coming into the switch, having the value 101110 within its header TOS field is assigned the priority 7 within priority field of the 802.1Q tag of the encapsulating frame coming out of the switch, if the out port is tagged.
In other word the incoming frame will be assigned to Class 7 of service using Level 3 marking, serviced via high priority queue and marked at Level 2 with priority 7.
The other switches should be aware of 802.1p priority using tagged ports for forwarding the frame.
If the out port is untagged the frame goes out of the switch unmarked, thus killing the prioritization.

2. dhcp-relay works only on routing switches, receves DHCP broadcast request and translates them into unicast or subnet directed broadcasts (depending on configuration), addressed to the specified DHCP servers or subnet.

3. Theoretically you're right practically it depends on segmentation of your network and traffic flows.

Good luck,
Dmitry