[e2e] MTU variability

RJ Atkinson rja at extremenetworks.com
Sun Aug 10 06:28:12 PDT 2003

On Sunday, Aug 10, 2003, at 06:28 America/Montreal, Lloyd Wood wrote:
> I think multihoming and handover the real cause of MTU variability.

Other sources of MTU variability that are increasingly common:
	- encrypted/authenticated tunnels using ESP/AH
	- frame-relay-like deployments of MPLS-based tunnels [1]
	- Gigabit Ethernet (and 10 GigE) with manually configured
		jumbo frame (9180 byte IP MTU) support [2]

Ran Atkinson
rja at extremenetworks.com

[1] Sometimes erroneously referred to as a Virtual Private Network
(VPN).  However, there is no security in this, hence no privacy,
hence it isn't really a "private" network.
[2] Officially, IEEE 802 doesn't admit that jumbo frames exist,
because an Ethernet frame much larger than 1518 bytes is incompatible
with Ethernet subnets that use CSMA/CD transmission.  However,
CSMA/CD is virtually non-existent with 1 GigE products.  CSMA/CD
is not permitted with 10 GigE.  So virtually all GigE (and 10 GigE)
products support a 9K byte MTU -- and have done for years now.

More information about the end2end-interest mailing list