Hi Tomas,
Sorry - didnt mean to keep you hanging this long. Work comes first
then fun stuff like this.
On Tue, 2004-01-27 at 06:59, Tomas Szepe wrote:
> On Jan-26 2004, Mon, 22:14 -0500
> jamal <hadi@xxxxxxxxxx> wrote:
[..]
> Thanks for your reply, Jamal. Unfortunately, we don't really understand
> your example. Please see below.
>
How about we take this offline. I help you and you document it so other
people understand it?
> What's the mechanism for matching the IPs? We need to insert
> thousands of these rules and shape constant 20+ Mbit flow of
> traffic. If it doesn't use a hash or similar, we're back to
> where we started.
Its the u32 classifier which is more sophisticated than the standard
iptables one. It does go into a tree of hashes. If you are unhappy with
it, it is trivial to add your own classifier as well.
[..]
>
> Would you know of any real documentation on tc/ingress that
> we could use to deconstruct this example and understand it?
>
Lets take it offline - there is no docs really. In particular this
feature i am talking about is not documented anywhere. Unfortunately
what that means is people reinvent it every summer. IMQ being one of
those inventions.
I would like to also satisfy Vladmirs requirement of delaying packets
instead of policing them.
cheers,
jamal
|