[nsp] ip load-sharing per-packet - cef accelerated ?

[Dan Hollis]

On Mon, 10 Mar 2003, Jared Mauch wrote:
> On Mon, Mar 10, 2003 at 02:22:12PM -0800, Dan Hollis wrote:
> > My understanding is that 'ip load-sharing per-packet' is CEF accelerated, 
> > yes?
> > Just trying to understand why some providers won't support per-packet CEF.
> 	A good reason they tend to avoid it is it causes
> out of order packets which result in lower end-to-end throughput.

We found the opposite. per-packet CEF on links where it is supported 
result in tremendous performance increase and smoother utilization of the 
links.

Maybe with super sucky tcp implementations, out of order packets is a 
problem. But we havent found it to be a problem.

Anyway, that's not why they claim they are avoiding per-packet CEF.

> 	What you want is a srcIP/dstIP/srcPT/dstPT based balancing
> to be perfectly honest.  This will keep the same flows on the same
> circuit.

No. That is not what we want.

The question is, does per-packet CEF cause router meltdown while 
per-destination CEF does not? Isnt the CEF lookup CPU utilization the same 
whether it's per-packet or per-destination?

-Dan
-- 
[-] Omae no subete no kichi wa ore no mono da. [-]