There is another theoretical limit, which is originating from how GRE
tunnels work.
Each packet in the GRE tunnel has a sequence number, and on fast links,
the counter warps around too fast...
That was the reason I got from Cisco for why they would not support GRE
on 12000...
Arie
> -----Original Message-----
> From: Hank Nussbacher [mailto:hank@att.net.il]
> Sent: Sunday, December 02, 2001 9:45 AM
> To: Brian Wallingford; cisco-nsp@puck.nether.net
> Subject: Re: [nsp] GRE tunnel bandwidth
>
>
> At 02:49 PM 11/30/2001 -0500, Brian Wallingford wrote:
> >Are there practical bandwidth limitations to GRE tunnels
> (Assume such
> >tunnels are restricted to a centrally-managed/owned network)? I was
> >unable to find anything definitive on CCO.
>
> We have run extensive tests on a 7206VXR, NPE400 sending to a
> 7513 w/ RSP4
> and VIP4s. The numbers change depending if you run with or
> without CEF and
> if you run with or without DCEF. Encap and decap also make a
> difference
> and well as with or without checksums. Short summary:
>
> - with no CEF and no DCEF, 22kpss with both the 7206 and the
> 7513 hit 99% CPU.
> - with CEF or DCEF at 106kpps, the 7206 (doing the encap)
> will hit 100% CPU
> and the 7513 will hit 96% CPU doing the decap.
> - with DCEF/CEF but doing no checksums, one can hit 107kpps
> on the 7206VXR
> doing encap and about 83% CPU and on the 7513 one will hit
> about 85% CPU
> doing the decap on the VIP.
>
> Regards,
> Hank Nussbacher
> Wanwall Ltd.
>
> >brian
>
>
>
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:57 EDT