Ethernet packet size and QinQ standard?

Have a question or want to start a discussion? Post it! No Registration Necessary.  Now with pictures!

Threaded View

A normative question:

AS we all know, ethernet has a max packet size of 1514 bytes:
14 bytes header (dst + src + len/type), plus 1500 bytes data.

For 802.1Q tagging, the spec was changed to allow a max packet size
of 1518 bytes: 1514 bytes, plus 4 bytes for the .1Q type/tag/prio fields.

For QinQ tagging, has the base spec been changed again,
now to a size of 1522 bytes, to allow for the double .1Q fields?
Are there any plans to go beyond this?

I'm kind of surprized because some hardware does have the packet size
hard-coded in, and while a careful study was done when the first
packet size change was made, changing this again and again may cause
trouble and I'd like to understand the rationale behind this.

(and no, jumbograms isn't the answer - suppose you have hardware which
is capable of 9kB packets, the same hardware may not be capable to carry
9kB+2bytes needed for .1Q, or 9kB+4bytes for QinQ protocols)


Geert Jan

Re: Ethernet packet size and QinQ standard?

On 08/12/09 15:06, Geert Jan de Groot wrote:
Quoted text here. Click to load it

The relevant standard is 802.1ad, aka "Amendment to IEEE 802.1Q-2005.
IEEE Standard for Local and Metropolitan Area Networks—Virtual Bridged
Local Area Networks—Revision—Amendment 4: Provider Bridges"

However, the maximum ethernet packet size was not changed from 1518 so
it is currently out of spec to provide an end-user MTU of 1500 using
Q-in-Q (as I understand it).

Work was started on 802.3as to expand the maximum frame size to 2000
bytes (support for 802.1ad being one of the objectives) but it seems to
have stalled. Given the vast install-base of ethernet equipment it seems
unlikely such a change would be worthwhile.

Re: Ethernet packet size and QinQ standard?


Quoted text here. Click to load it

there is probably a difference between the minimum constraint the
standards require and the stuff deployed in some networks.

However adding more 4 byte "wrappers" in an Ethernet frame is becoming
common - at least in a carrier environment.

every MPLS label adds 4 bytes, so for a typical carrier MPLS VPN
service you need 2 levels, with more for various add ons such as fast

At work the switches are set to allow 3 or 4 levels of MPLS label -
and they flow through the L2 Ethernet switches that glue everything
together OK (Cisco with non default max size set).

Also the Ethernet access tails deployed for MPLS use Q in Q.
801.q customer traffic is sent across a "multi service" bearer, where
an extra layer of 802.1q tags are used to segregate different services
from each other + management.

then there is all the malarky about large packet - where the actual
maximum sze depends on the vendor + equipment + blade type....
Regards - replace xyz with ntl

Re: Ethernet packet size and QinQ standard?

no it hasnt stalled, 802.3as was ratified. I am pretty sure you'll find it
included in 802.3-2008.

The frame length was increased to 2000 but the payload remains the same at
1500. The extra is to allow for every sort of encapsulation that may be
needed in the future

Quoted text here. Click to load it

Re: Ethernet packet size and QinQ standard?

Quoted text here. Click to load it

The more things change, the more they stay the same.

The original 1500 byte Ethernet frame was *intended* to carry 1K bytes
of payload. The additional 476 bytes was for "every sort of
encapsulation that may be needed in the future." Of course, folks took
the entire field for the payload, and so 28 years later we had to
expand the frame for the encapsulation.

Let's see if we have to do it again in 2036.

Rich Seifert              Networks and Communications Consulting
                                 21885 Bear Creek Way
(408) 395-5700        Los Gatos, CA 95033
(408) 228-0803 FAX

Send replies to: usenet at richseifert dot com

Site Timeline