Hello Ajain
In order to address your first question, I labbed this up and found that the IP MTU of the subinterface is only limited by the interface MTU of the parent interface. It is otherwise independent of the IP MTU configured on the parent interface. In the following example, I have set the MTU of the parent interface to 1100, and the IP MTU of the parent interface to 1050. Take a look at what range I am allowed to configure the IP MTU of the subinterface:
R1(config)#inter gig 0/0.1
R1(config-subif)#ip mtu ?
<68-1100> MTU (bytes)
The subinterface IP MTU is therefore limited by the interface MTU of the parent interface.
Now as for the interface MTU, it is possible to modify this on the subinterface, but the router will always take a look at the interface MTU of the parent interface and the subinterface, and use the smaller of the two.
For your second question concerning Etherchannel, I assume you don’t mean subinterfaces here, because you cannot use Etherchannel in conjunction with subinterfaces. I assume you want to know how the individual physical interfaces and the port-channel interface interact as far as MTU goes.
For Layer 2 Etherchannel, only the interface MTU can be adjusted. You must change only the MTU of the port-channel. Any change to that MTU will automatically change the MTU of all of the physical interfaces, if they support the configured MTU. If you are using a switch that does not support individual MTU configurations per port, you must use the system mtu
configuration, which will change the MTU on all ports of the switch.
For Layer 3 Etherchannel, the individual port IP MTUs should not be modified. Only the IP MTU of the port-channel should be modified. If you modify the IP MTU of individual ports, the IP MTU of the port-channel will inherit the smallest IP MTU configured on any of the physical ports. This will cause problems on your link as some physical channels may fragment IP packets or may drop them.
For the third question, it’s not possible to define a maximum segment size for UDP as you would with TCP. THis is because TCP has negotiation mechanisms between sender and reciever. When there is a perceived bottleneck somewhere on the network (which can be artificially created using the MSS feature on a network device), the sender and receiver will actually renegotiate the size of the segment. UDP has no such negotiation, so this cannot be done.
I hope this has been helpful!
Laz