Fragmentation of MPLS Packets

Unanswered Question
Oct 12th, 2009
User Badges:


We have several banking branch. All of them connect to Data Centre with MPLS links via GRE Tunnel and for Backup connection; they connect to DC via WiMAX. We should set MTU 1400 on these MPLS and WiMAX interfaces because of the nature of MPLS & WiMAX. If we use TCP protocol for data transmission, we should set the DF bit ZERO. But the problem happens when we want to use UDP protocol with packet size greater than 1500 bytes for some of our applications.

these packets can't be reassemble via servers and we are forced to set virtual reassembly command in all of MPLS and WiMAX links, but I don't think it could be the best solution regarding to Routers load and cpu usage.

Could you pls help me according to this subject.

Best Regards

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Giuseppe Larosa Tue, 10/13/2009 - 01:24
User Badges:
  • Super Silver, 17500 points or more
  • Hall of Fame,

    Founding Member

Hello Ladan,

IP fragmentation and reassembly is provided by IPv4 header using specific fields.

I don't understand why those servers cannot reassembly UDP PDUs and they look like to be able to reassembly TCP PDUs.

I think this may come from some security fixes on server operating system.

MPLS frames cannot be fragmented. However when carried inside a GRE packet that is an IPv4 packet, fragmentation can occur using services of external IPv4 header.

I would say:

or applications are redesigned to not use UDP packets greater then X bytes

or server operating system is tuned to accept to reassembly UDP packets.

Clearly performing reassembly on routers lead to a performance penalty for them.

Hope to help



This Discussion