[Xorp-users] Why Xorp so slow ?

[Bruce Simpson]

Hi,

Thanks for your message about XORP performance.

There are reasons for this performance difference, and perhaps the most 
crucial is that XORP uses an intermediate RIB process; OpenBGPD doesn't, 
Quagga does do something like a RIB. The projects you're comparing also 
have vastly different goals.

Sébastien Namèche wrote:
> ...
> It's old hardware, in production we will use servers 2 ou 3 times  
> faster. But still, I don't understand why there is so much differences  
> between Xorp and the others.
>   

What exactly is the specification of the hardware you're using, to give 
us a better feel for the slowdown? Did you perform these comparisons on 
the same hardware?

    The thing to be borne in mind is that for many years, XORP's goal 
was correctness and stability, not necessarily performance, and the 
design was aimed at non-embedded platforms in mind initially. So 
premature optimization wasn't a goal for the project.
    It is also worth pointing out that as a result of this focus, XORP 
hasn't suffered from issues in production deployment which have bitten 
the alternatives.

    Of course in using these other tools you make a trade-off -- the 
other alternatives don't support multicast routing, for example, or 
necessarily run cross-platform, or support all of the policy 
redistribution features, MANET protocols etc.

The danger in informal benchmarks is that it's easy to trip up over 
things like the L2 cache being primed, kernel memory fragmentation, etc.

Do you plan to deploy XORP? We can get performance up, however, it's 
also a community effort at the end of the day.

thanks
BMS