Path MTU Discovery
Page 25-42

Path MTU Discovery

All Gigabit Ethernet modules and all Mammoth-based Ethernet modules on the Omni
Switch/Router in Release 4.0 and later support path Maximum Transmission Unit (MTU)
discovery. In path MTU discovery, the Ethernet frame (datagram) size is set to the largest size
that does not require fragmentation anywhere along the path from a source host to its desti-
nation. This frame size, known as a Path MTU (PMTU), is thus equal to the minimum of the
MTUs of each hop in the path.
♦ Note ♦
MTU discovery is not supported on token ring, FDDI,
WAN, or non-Mammoth Ethernet modules. However,
token ring and FDDI can be used as intermediate links
(e.g., trunking or bridging) between remote switches.
Path MTU discovery is active all of the time and is part of the switch’s operating system; you
do not need configure it.
The source host initially assumes that the PMTU of a path is the MTU of the first hop. It sends
all datagrams with the “Don’t Fragment” (DF) bit set. If a switch/router along the path receives
a datagram that is too large to forward without fragmentation, the following steps will be
executed:
1. The switch/router that cannot forward these datagrams (i.e., the constricting hop) will
discard them.
2. The constricting hop will send ICMP destination unreachable messages to the source host
with a code that indicates fragmentation is needed and the “Don’t Fragment” (DF) bit in
the Internet Protocol (IP) header has been set. This message (known as a “Datagram Too
Big” message) contains the PMTU of the constricting hop.
3. After receiving a “Datagram Too Big” message, the source host reduces the size of the
MTU so it matches the PMTU of the constricting hop.
4. The MTU discovery process ends when datagrams can be sent without fragmentation.
However, the source host will not reduce the size of a datagram below 68 octets.