[c-nsp] c7206VXR(NPE-G1) w/ 12.2(18)S8 - Memory failures

[Rodney Dunn]

I can tell you up front that the developers are very
aware of memory resource issues and try extremely hard to use
as little as possible.  I can see how from the outside it
may look like it's some ploy to force upgrades. I can tell
you that is not true.

As for the 12.2(25)S changes that came out where some
infrastructure changes were made to the CEF code it ended
up resulting in more memory usage than the previous changes.

Work was done to bring the memory usage back to the order
it was before those changes were put in. Now it's a matter
of getting those images in the field.

In a perfect world it would have been nice if the first release
out the door would have been on par with the prior release in this
area but it wasn't.

I asked about this specific issue of when this code would be
available on CCO that brings the memory usage per prefix back
down and the target date is the fall of this year.

This doesn't account for more and more features though.
Memory usage by those is unavoidable.

I checked a 12.0(27)S3 vs. a 12.2(25)S3 image for the 75xx
yesterday and the decompressed memory usage was around 4M.
 
Rodney



On Thu, Mar 31, 2005 at 01:34:46AM -0800, Ted Mittelstaedt wrote:
> cisco-nsp-bounces at puck.nether.net wrote:
> > On Tue, Mar 29, 2005 at 10:34:39AM +0200, Santiago Felipe Perez wrote:
> >
> >
> > 	I would upgrade memory to 512m or even 1G as if you upgrade
> > to 12.2(25)S in the future you will see significant memory usage.
> >
> 
> The NPE 300 and NPE 225 only go to 256MB of ram.  What is your suggestion
> there?
> 
> > 	I recommend that anyone who is running full bgp routes upgrade
> > to at least 512m, if not more if your system can handle more
> > just to reduce
> > the impact of any memory leak bug or other problem with the system.
> >
> 
> Not too long ago 128MB was sufficient for a full BGP table.  Then it went
> to 256MB which killed use of all NPE's under NPE225.
> 
> Now it is at 512 which kills all NPE's save the G1.
> 
> Funny thing though - I don't recall the global BGP table doubling is
> size when 128MB of ram became too small, nor do I recall it doubling in
> size when 256MB became too small.
> 
> It is sad to see that Cisco's developers have been told to force
> obsolescense
> on the customers.
> 
> Ted
> 
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/