[c-nsp] CEF memory usage

[Rodney Dunn]

On Fri, Oct 07, 2005 at 10:13:56AM -0500, Pete Templin wrote:
> Rodney Dunn wrote:
> 
> > The overall throughput improvement of a box with multiple features
> > applied on various interfaces improved with the CEF changes also.
> > How much depends on the traffic profile/distribution through the
> > interfaces and the features configured. It was an improvement
> > in overall packet switching performance. It's very likely those
> > changes will show in i a future T release at some point such
> > as 12.4T.
> 
> Are you saying that the 12.2S improvements will merge back into T 

Not generally. We may hand pick a couple depending on need.
The CEF switching vector improvements are one such that may very
well end up on the radar list.

(and 
> subsequently mainline),

No. Nearly impossible to get new features in Mainline. They have to go
in T and let that turn in to the next mainline.

> or that other improvements will appear in 12.4T? 

There are new features/improvements going in to T code all the time.
That's the purpose of it.


>   12.4T won't be available for 7500, so it's an "other improvement" I do 
> hope it loops back to an 12.xS(y) release at some point.

The future features will be in 12.2SB. What exist in 12.4 for the
75xx will be there but new features moving forward will not.

That doesn't mean every single 12.4T feature will be in 12.2SB.

The major general features we will try and keep parity between
the paths.


> 
> pt