RE: [bmwg] Minutes for the bmwg session at IETF-69

["Chip Popoviciu \(cpopovic\)" <cpopovic at cisco.com>]

David,

It seems I interpreted Dan's earlier comments differently. Nevertheless,
the points you made below (and Tom completed) are reasonable. So let me
summarize the conclusion of the discussion then:

1) The recommended frame size set stays as is (RFC2544) 
2) Will add a note such as:

"Note: The recommended 1518 bytes frame size represents the maximum size
of an untagged Ethernet frame. According to the IEEE 802.3as standard,
the frame size can be increased up to 2048 bytes to accommodate frame
tags."

This text simply indicates to options available without elaborating on
benchmarking with tagged frames which is outside the scope of the
document. A tester can choose to use VLAN tags and naturally account for
them at all frame sizes (similar to testing with any type of additional
tags or headers). The note simply helps the reader understand the upper
bound options. Please provide comments on the write-up.

Best Regards,
Chip

-----Original Message-----
From: David Newman [mailto:dnewman at networktest.com] 
Sent: Friday, August 03, 2007 5:46 PM
To: bmwg at ietf.org
Subject: Re: [bmwg] Minutes for the bmwg session at IETF-69

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 8/3/07 2:32 PM, Chip Popoviciu (cpopovic) wrote:
> It seems we have an agreement to include 1522 and 2048 in the main
set.

I do not agree that *any* +4 VLAN length such as 1522 belongs in the
main set. Dan's comment as AD seemed to support my suggestion of having
a separate section for testing with VLANs.

This is also conceptually cleaner; there's one standard set of lengths,
and then this set optionally can be modified for users interested in
VLAN testing. Not everyone uses VLANs, and some folks that do use
multiple VLAN tags. A separate section allows the use of one or more
tags without forcing those lengths on everyone else.

> Also add a note that these two frame sizes reflect the evolution of
the
> 802.3 standard to include VLAN tagging.   
> 
> I am still not sure if we need to keep 1518. I acknowledge the point 
> made that it might be good to have all the previously recommended 
> sizes for data comparison however, I am not convinced it is necessary.

It's not just for comparison with RFC 2544 and 2889 and 3918 results,
though that is very meaningful for anyone trying to compare IPv4 and
IPv6 performance, or to conduct mixed-version tests.

The other issue is that 1518 still is a maximum length for Ethernet, as
given in IEEE 802.3-2002. My understanding (possibly erroneous, and I
hope Dan or someone else will correct me if so) is that 802.3as
maintains the same frame format and data length field as previous
versions.

Ergo, a tester not hobbled by QinQ and wrapped in ESP and AH over a GRE
tunnel etc. etc. etc. will still have a maximum length of 1518 bytes.

dn

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.3 (Darwin)

iD8DBQFGs6IyyPxGVjntI4IRAgYvAJ41HNucOOri0Zwwr6NIYF257SVo8gCg65ax
H/Jr/rFo6z6qynE2WAyb/RU=
=BgKN
-----END PGP SIGNATURE-----

_______________________________________________
bmwg mailing list
bmwg at ietf.org
https://www1.ietf.org/mailman/listinfo/bmwg

_______________________________________________
bmwg mailing list
bmwg at ietf.org
https://www1.ietf.org/mailman/listinfo/bmwg