[AusNOG] Why not Symmetric ingress and egress?

[Chris Pollock]

>From what I remember, there's an incredible number of settings for
setsockopt, many more than are listed on that page.

This page makes reference to being able to set the receive window size, but
the link it points to doesn't list what the parameter is called.
http://msdn.microsoft.com/en-us/library/ms819736.aspx

Or you could just do what too many Windows developers do - require your
application to be run with Administrative privileges and change the registry
setting manually :P
--
Chris Pollock
::loss4words::versatile solutions for modern living::
w: http://loss4words.com
e: chris at loss4words.com
m: 0410 747 765
f: 07 3129 2351


On Sat, Jun 19, 2010 at 8:33 PM, Dmitri Kalintsev <dek735 at gmail.com> wrote:

> Chris,
>
> Thanks for that. The idea that an application could potentially work around
> system-wide settings is intriguing, but the description of setsockopt() does
> not seem to indicate the possibility to set send and/or receive windows
> size, which are the parameters in question:
>
> http://msdn.microsoft.com/en-us/library/ms740476%28VS.85%29.aspx
>
> I would gladly stand corrected, as this is a very interesting and relevant
> question. The reason why it is, is because the topic of performance of TCP
> applications over "long fat pipe" type of transmission infrastructure keeps
> on coming up from customers I land up dealing with (indirectly) in a pretty
> persistent fashion.
>
> On Sat, Jun 19, 2010 at 8:17 PM, Chris Pollock <youare at loss4words.com>wrote:
>
>> Handled on the OS level by default, yes.  The application can do something
>> about it though, using the "setssockopt" function, which sets it on a
>> per-socket basis IIRC.
>
>
> Regards,
>
> -- D
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ausnog.net/pipermail/ausnog/attachments/20100619/4f4a3898/attachment.html>