[nsp] Performance impact of IP-in-IP tunnels

[Volodymyr Yakovenko]

On Wed, Jun 18, 2003 at 04:49:03PM -0400, Rajesh Talpade wrote:
>Hi
>
>Pointers to data on the performance impact of sending packets thru an
>IP-in-IP tunnel will be much appreciated. Examples of data ...
>
>- what is maximum thruput for a router with packets going out of a tunnel
>interface, compared to a non-tunneled interface?
>- what is the impact on non-tunneled traffic at a router when there is 
>significant traffic going out of a tunnel interface?

http://www.cisco.com/en/US/products/sw/iosswrel/ps1828/products_configuration_guide_chapter09186a0080087da4.html

* Encapsulation and decapsulation at the tunnel end points are slow operations;
  in general, only processor switching is supported. However, fast switching of
  GRE tunnels was introduced in Cisco IOS Release 11.1 for the Cisco 2500 
  series and the Cisco 4000 series of routers.

I believe GRE fast (and CEF) switching is supported on the most of Cisco 
router's hardware architectures.

If IP-in-IP tunnels are also switched by interrupt handler (fast switched) they
should have small effect on system performance.

-- 
Regards,
Volodymyr.