I am currently studying for the ISCW exam and have a question regarding MTU. When studying GRE or IPsec tunnels the MTU size can be reduced to support links that do no support the additional overhead of the encapsulation through out the topology. This made perfect sense to me so things don't have to be fragmented and retransmitted.
Then I started reading the MPLS chapter and it stated with the additional overhead of the MPLS frame you could increase the mpls mtu size. So for example on an ethernet link it could go from 1500 to 1504.
Why would it be increased if there could be potential fragmentation and MTU issues with other links in the topology?
If you increased the MTU for MPLS this will only have an effect within the MPLS network. Once it has left the MPLS network there is no label and so you don't need a larger mtu. So you send a normal ethernet frame of 1500. It gets to the first MPLS router/switch which sets the MTU to 1504, adds a label and switches it to the next MPLS device. This device also has the MTU size set to 1504 so no fragmentation is needed. When the frame is sent from the egress MPLS device to a non-MPLS device there is no label, because the label is only used within the MPLS network and so the frame is 1500 again.
So there is not fragmentation anywhere within the MPLS network.
" if the provider was provided an MPLS VPN wan solution would they change their MTU accordingly"
not quite sure what you are asking here. If a customer purchased an MPLS VPN solution from a provider then the customer would not have to change the MTU on any of their devices because none of the customer links would be running MPLS.
The MTU only becomes important within the provider MPLS network. And as previously stated the MPLS network is under the administrative control of the provider so they could set the MTU wherever they needed to.
Has this helped ?
the MPLS provider has to verify that a customer IP packet of size 1500 bytes can travel between MPLS VPN sites.
This requires at least an MPLS mtu of 1508 on routers, to be noted also all L2 LAN switches in the middle need to have their L2 MTU increased to support the additional overhead.
the MTU test with extended ping of 1500 bytes with DF set (or not set) between sites is a standard test when deploying a MPLS L3 VPN service.
the mpls mtu as ip mtu is not dependent on the type of link (no dependency from GE or POS or other types of L2 technologies in use).
Think of MPLS as Layer 2.5 in the OSI stack in the forwarding plane.
MPLS relies on IPv4 for the signalling plane (label creation and distribution)
Hope to help