Re: [mpls-linux-general] really - really need help

[Yon U. <hav...@gm...>]

Hi,

On Tue, 28 May 2002, venda y wrote:
[snip]

> So, can anyone told us how to show that MPLS is better than IP ?

 I'll assume Better == faster forwarding.

Why? What makes you think so?

 In theory looking up a label could be *implemented* faster than looking
up a route. But, repeating myself, it is *implementation* dependant.
Actual implementations (for example network processors) will have
different delays for each operation, including the faster, slower and
equal delay case, and that might change for different network
element states and input conditions (as you are actully testing).

 In theory mpls-linux could implement switching a mpls packet in way
that makes that operation faster than vanilla linux ip routing. In
practice you just measured that this is not the case for all packet
sizes. The performance of the whole system is dependant on other factors,
like network cards/drivers (and interrupt handling (NAPI), which
*could* be protocol independent).

> what
> make mpls-linux different than IP (or better than IP) ?

 The forwarding path is not (necessarily) bound to the IP forwarding path.
There are implicit (LSP)-tunnels allowing new network features (l2&l3
vpn, etc.).

> Because our
> purpose to work on it is to prove that MPLS is better than IP

 Define better (in more than one word ("faster" (faster *what*?, under
what conditions?))).

SCNR
 yon