Marcus,
Looking at the article on Jumbo frames
that you referenced at http://sd.wareonearth.com/~phil/jumbo.html,
it includes the statement:
“while the number of packets larger than 1500 bytes appears small, more than 50%
of the bytes were carried by such
packets because of their larger size.”
However, when you look at the source of
this information (reference 1), what it actually says is:
“over half of the bytes are carried
in packets of size 1500 bytes or larger”
This is, of course, totally different.
As the red curve of figure 3b of http://www.caida.org/publications/papers/1998/Inet98/Inet98.html
shows, it is the 1500 byte packets that contain ~ 55% of the bytes in the
analysis and in fact the total number of bytes in all packets above 1500 is a negligible
percentage of the total.
Regards,
Pete Anslow
Nortel Networks UK Limited, London Rd, Harlow, Essex CM17 9NA,
UK
External +44 1279 402540 Fax +44 1279
405670 ESN 742 2540
Email: pja@xxxxxxxxxx
From: Marcus Duelk [mailto:duelk@xxxxxxxxxx]
Sent: 10 August 2006 16:21
To:
STDS-802-3-HSSG@xxxxxxxxxxxxxxxxx
Subject: Re: [HSSG] Topics for
Consideration: Jumbo Frames
Hi Arthur,
there is a nice little "tutorial" about the Pros and Cons
of jumbo frames at http://sd.wareonearth.com/~phil/jumbo.html,
it also includes actual frame size distribution measurement on the
MCI backbone from 1998 in which you can see that there are a lot
of frames >1500B from FDDI and other sources, they mention that
more than 50% of all traffic (measured in Bytes and not in packets)
come from packets >1500B. But, larger packets also have a clear
benefit to TCP/IP performance. The difference in store-and-forward
latency between 1.5kB and 9kB is negligible at higher speeds, time
of flight is much, much higher ...
Marcus
Arthur Marris wrote:
Joel, Mike,
Don't jumbo frames increase
latency in links where there is store and forward?
Arthur.
Both my position and that of the customers I deal with is that higher speeds,
10gbps included, is more efficient with the larger packets.
no argument from me on this one, although the actual
efficiency gained comparing 1.5 K frames to 9 K frames is only ~ 2%, but when
you're on the cutting/bleeding edge of technology, every bit counts.
--
___________________________
Marcus Duelk
Bell Labs / Lucent Technologies
Data Optical Networks Research
Crawford Hill HOH R-237
791 Holmdel-Keyport Road
Holmdel, NJ 07733, USA
fon +1 (732) 888-7086
fax +1 (732) 888-7074