[Xorp-users] Multicore scalability

[Bruce Simpson]

Jaime Nebrera wrote:
> ...
>   Yep, I'm aware of this. Actually thats why I was making my question. I
> have been reading quite a lot on forwarding improvements (RSS, lock
> contention reduction, flow affinity, bottlenecks, ...) but in this docs
> they dont talk about XORP side by itself. As I'm aware XORP is
> monothread per routing protocol, I didnt want to invest long time in
> optimizing the forwarding plane and then discover that doesnt work
> properly with XORP :(
>   

Well, XORP is control plane software. Sure, it could benefit from 
optimization of the FEA->kernel plumb -- what wouldn't -- normally the 
FEA batch loads this stuff, at the end of the day it's limited by 
whatever interface is available to the forwarding plane.

Most of the time BSD/Linux are not being used in the Default-Free Zone, 
so the defaults are not optimized for that use case. It entirely depends 
on what one wishes to achieve.

All software forwarding planes have similar limitations, the hardware 
solution is usually some form of TCAM (ternary content-addressable 
memory). Various platforms (e.g. in Juniper, Cisco, other vendor space) 
have more or less TCAM.  It is worth bearing in mind which use case of 
the router you are optimizing for -- routers in the Default-Free Zone 
have quite different performance requirements from edge routers or MPLS 
nodes.

In fact one could say that was one big motivation behind MPLS... to save 
money on expensive TCAM boxes.

cheers,
BMS