About MLS evolution.

[speculor_cisco]

I know that MLS began as a dual effort between a Route Processor and a Switch Engine.

Take a look to the following evolution:

Catalyst 5000: external router without shortcut.

Catalyst 5000 + RSM: internal router without shortcut.

Catalyst 5000 + NFFC: external router with shortcut.

Catalyst 5000 + RSM + NFFC: internal router with shortcut.

where:

RSM: Route Processor.

NFFC: Switch Engine.

Is it correct?

I imagine that there is also the Supervisor in the switch yet.

So, I can say that, in the last case, there are the following items:

Supervisor: the "old" intelligence which manages the switch fabric.

NFFC: the "new" intelligence, the Switch Engine, which manages the shortcut.

RSM: the "new" intelligence, the Route Processor, which manages routing.

Is it correct too?

[Jon Marshall]

speculor_cisco wrote:

I know that MLS began as a dual effort between a Route Processor and a Switch Engine.

Take a look to the following evolution:

Catalyst 5000: external router without shortcut.

Catalyst 5000 + RSM: internal router without shortcut.

Catalyst 5000 + NFFC: external router with shortcut.

Catalyst 5000 + RSM + NFFC: internal router with shortcut.

where:

RSM: Route Processor.

NFFC: Switch Engine.

Is it correct?

I imagine that there is also the Supervisor in the switch yet.

So, I can say that, in the last case, there are the following items:

Supervisor: the "old" intelligence which manages the switch fabric.

NFFC: the "new" intelligence, the Switch Engine, which manages the shortcut.

RSM: the "new" intelligence, the Route Processor, which manages routing.

Is it correct too?

Things have changed quite a bit since the 5500 switch.

Just a correction to the above before comparing with 6500. The NFFC is like the PFC on a 6500 ie. it could switch traffic both L2 and L3 but it is not a router. Without an RSM the 5500 was still only a L2 switch. You have to have an RSM for it to be L3 even though the RSM didn't actually switch the traffic. Even then it was still only non CEF MLS switching in that the first packet had to be sent to the RSM for the L3 lookup.

Compare this with the 6500 now which does CEF based MLS.

The MSFC is the equivalent of the old RSM, difference is due to CEF the first packet does not necessarily have to be sent the MSFC.

The PFC is the equivalent of the NFFC ie. it switches traffic both at layer 2 and layer 3. The PFC also does a lot more such a QOS implementation but i'm not sure whether the NFFC did the same.

The supervisor is simply the card that contains the PFC and the MSFC.

In a modern 6500 the MSFC is responsible for the control plane functions, one of the primary ones being establishing the routing table by peerings with neighbors etc. It builds the RIB (Routing Information Base)

The PFC is responsible for the data plane functions ie. the actual forwarding of data both L2 and L3 traffic as well as various other functions. The RIB built by the MSFC is downloaded to the PFC to create a FIB (Forwarding Information Base).

So without an RSM/MSFC no RIB is built and so the NFFC/PFC cannot do L3 switching.

Jon

[speculor_cisco]

Hello Jon and thanks for your answer.

You wrote that the supervisor is simply the card that contains the PFC and the MSFC.

Can I say that the supervisor is a type of mother board?

In this case, are PFC and MSFC separate modules or integrated?

Can you say me some documents which can help me about these items?

Thanks.

[Jon Marshall]

speculor_cisco wrote:
Hello Jon and thanks for your answer.
You wrote that the supervisor is simply the card that contains the PFC and the MSFC.
Can I say that the supervisor is a type of mother board?
In this case, are PFC and MSFC separate modules or integrated?
Can you say me some documents which can help me about these items?
Thanks.

Have a look at this architecture paper on the 6500. It has a section on the supervisor and the cards intergrated into the supervisor -

http://www.cisco.com/en/US/prod/collateral/switches/ps5718/ps708/prod_white_paper0900aecd80673385.html

Jon

[Peter Paluch]

Jon,

Wonderful answers. I did not yet have the time to acquaint myself with the MLS as it once was, and while reading your first reply here I got totally lost among the acronyms Nevertheless, I have a better understanding now of the stuff than I had before - and I hope to be able to ask you further.

Best regards,

Peter

[Jon Marshall]

paluchpeter wrote:
Jon,
Wonderful answers. I did not yet have the time to acquaint myself with the MLS as it once was, and while reading your first reply here I got totally lost among the acronyms  Nevertheless, I have a better understanding now of the stuff than I had before - and I hope to be able to ask you further.
Best regards,
Peter

Peter

Many thanks for kind words and ratings.

Yes MLS/NFFC/RSM/PFC/MSFC - it is a bit acronym mad isn't it. Just in case you're interested -

MLS = Multi layer switching, suspect you knew that one !

NFFC = Netflow Feature Card

RSM = Route Switch Module

PFC = Policy Feature Card

MSFC = Multilayer Switch Feature Card

Jon

[speculor_cisco]

Jon, thanks for the document.

It is very interesting and I will spend time with it.

In the CCNP Official Certification Guide you can read:

"The actual MLS process can take two forms: interVLAN routing and Cisco Express Forwarding (CEF)."

Why does the author make this strong difference?

Does the author mean interVLAN routing without "route once switch many" or CEF?

[Jon Marshall]

speculor_cisco wrote:
Jon, thanks for the document.
It is very interesting and I will spend time with it.
In the CCNP Official Certification Guide you can read:
"The actual MLS process can take two forms: interVLAN routing and Cisco Express Forwarding (CEF)."
Why does the author make this strong difference?
Does the author mean interVLAN routing without "route once switch many" or CEF?

Hmm, not sure what the author is getting at at there to be honest. MLS or CEF based MLS, which is the modern variant, is how a L3 switch forwards L3 traffic. So i'm not sure how this relates to inter-vlan routiing unless the author is simply saying inter-vlan routing involves MLS. Or unless he is saying CEF is L2 and inter-vlan routing is L3 but again not entirely convinced.

Perhaps more context is needed from the book as it's difficult to say based on that one sentence.

Jon

[speculor_cisco]

I have added an image from the book.

If there was been only the first two scenarios, then that phrase would have made sense.

But with the third scenario, MLS is doing interVLAN routing and, in his logic, without CEF.

The only way to do that is "route once switch many".

But when the author begins to speak about CEF, first he considers the history of MLS

with "route one switch many" before and CEF after.

So, that third scenario has no explanation in my opinion.

As considering interVLAN routing the author introduces SVIs, do you know if I can configure SVIs

also with the two types of MLS? I would say it is clearly possible and I am quite sure

I have already seen something like this.

[Jon Marshall]

speculor_cisco wrote:
I have added an image from the book.
If there was been only the first two scenarios, then that phrase would have made sense.
But with the third scenario, MLS is doing interVLAN routing and, in his logic, without CEF.
The only way to do that is "route once switch many".
But when the author begins to speak about CEF, first he considers the history of MLS
with "route one switch many" before and CEF after.
So, that third scenario has no explanation in my opinion.
As considering interVLAN routing the author introduces SVIs, do you know if I can configure SVIs
also with the two types of MLS? I would say it is clearly possible and I am quite sure
I have already seen something like this.

Will have a read when i get a moment.

Quick answer to your question about SVI.

SVI = Switched Virtual Interface and yes you can configure them. In fact this is generally how you setup routed interfaces on a L3 switch eg.

int vlan 10  <-- this is an SVI

ip addres 192.168.5.1 255.255.255.0

Jon

[speculor_cisco]

Jon, thanks for your patience.