Tunnel overhead [On killing IPv6 transition mechanisms]

Gert Doering gert at space.net
Wed Mar 17 20:10:14 CET 2010


On Wed, Mar 17, 2010 at 11:55:46AM -0700, Ted Mittelstaedt wrote:
> I am pretty sure I recall reading that the larger routers have the
> ability to lower priority of ICMP echo replies and that most of them
> do.  

They do, which is why we ping servers, not routers :-)

> "...Network latency is without question on that list....additional 
> client-server latency due to IPv6 being used will quickly add
> up to seconds in terms of overall page load time for a complex web site..."
> That is factually incorrect.  TCP/IP uses sliding windows.  As long
> as the overall latency falls within the window setting, (and 150ms
> definitely does) then the IP transmission WILL NOT fall back into
> the "send-a-pack-wait-for-an-ack" mode and bandwidth will NOT be
> affected, thus it does not result in any increase in page load
> delay for larger pages.

A web page consists of manymanymany round-trips for manymanymany small
page elements.  150ms adds up.  It's worse with HTTP persistant 
connections, less bad with multiple parallel connections.

Web is not "FTP download a single big file in a continuous mostly
unidirectional stream".

Gert Doering
        -- NetMaster
Total number of prefixes smaller than registry allocations:  150584

SpaceNet AG                        Vorstand: Sebastian v. Bomhard
Joseph-Dollinger-Bogen 14          Aufsichtsratsvors.: A. Grundner-Culemann
D-80807 Muenchen                   HRB: 136055 (AG Muenchen)
Tel: +49 (89) 32356-444            USt-IdNr.: DE813185279
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 306 bytes
Desc: not available
Url : http://lists.cluenet.de/pipermail/ipv6-ops/attachments/20100317/adab4360/attachment.bin 

More information about the ipv6-ops mailing list