Multilayer Switching Route Processing / Service Engine

Joshua Davis · ‎04-14-2010

Ok I really hope that someone can help.

I am reading and trying to process the cisco press article:

Understanding and Configuring Multilayer Switching

http://www.ciscopress.com/articles/article.asp?p=700137

Am I correct in my understanding that MLS SE is the 1st generation MLS equivalent of MLS RP which seems to be capable of performing packet switching AND access-lists, QoS in hardware whereas MLS SE seems to be only capable of CRC, Source and Destination MAC address changes in hardware.

Does CEF use MLS SE since it doesn't function within the Route Caching 1st generation model or does it only use MLS RP?

Any help GREATLY appreciated.

Thanks

Joshua

Giuseppe Larosa · ‎04-14-2010

Hello Joshua,

CEF has a totally different approach:

CEF is topology driven, that is if there are 1000 routes in IP routing table CEF builds 1000 entries for these prefixes.

CEF uses also a neighbor table (adjacency table) for better efficiency.

Older method was flow based: first packet had to go to the route processor, the multilayer switching engine builded a temporary entry and it will be ready to capture the packet after it has been processed by route processor (packet rewrite). A complete entry for MLS is built and following packets are processed like the first one (same packet rewrite that means: same outgoing vlan-id if travels on a trunk, same source MAC address (of RP) and same destination MAC address (final destination or RP next-hop MAC address).

With CEF the first packet of a flow for a destination is processed by multilayer switching as the following ones.

So the ACLs need to be implemented on the multilayer switching engine there is no chance to have first packet processed by RP anymore.

For example: in a C6500 there are multiple brains: the L2 supervisor, the L3 MSFC, the PFC that is the multilayer switch and that implements also QoS.

What we miss with CEF is the capability to use an external router as RP that was possible with flow based MLS.

Hope to help

Giuseppe

Joshua Davis · ‎04-14-2010

Argh... this is so confusing.

Please see this link:

http://www.cisco.com/en/US/docs/ios/12_0/switch/configuration/guide/xcmls.html

It states that:

•Multilayer Switching-Switching Engine (MLS-SE)—A NetFlow Feature Card (NFFC)-equipped Catalyst 5000 series switch.

•Multilayer Switching-Route Processor (MLS-RP)—A Cisco router with MLS enabled.

•Multilayer Switching Protocol (MLSP)—The protocol running between the MLS-SE and MLS-RP to enable MLS.

Ok so I thought NetFlow was part of Traditional MLS (generation 1 in the CCNP BCMSN Official Exam Certification Guide 4th edition)

http://www.ciscopress.com/articles/article.asp?p=700137

So it seems as if it works like this.

Routing protocol populates Routing Table

Access lists and Routing table entries are then provided by MLS-RP to MLS-SE so that these packets can be layer 3 switched rather than layer 3 routed in hardware.

So in CEF no packet is actually being routed by the routing table but all packets are being forwarded via the prepopulated FIB correct?

So even the first packet in a flow of packets gets switched and not routed because the entire routing table is already in CEF.

the The Route Switch Module or RSM seems to be the module that actually runs the MLSP to move MLS-RP entries to MLS-SE entries.

So MLS-RP entries are they in hardware in ASICs or is that software based?

Admittedly I haven't spent more than 30 minutes trying to understand this and it seems like more time than that is required but your help would be greatly appeciated.

Joshua

Jon Marshall · ‎04-14-2010

Joshua

So in CEF no packet is actually being routed by the routing table but all packets are being forwarded via the prepopulated FIB correct?

So even the first packet in a flow of packets gets switched and not routed because the entire routing table is already in CEF.

Correct, because CEF is not dependant on the entry being built when the first packet in a flow arrives, it is already built and therefore there is no need to send the first packet to the route processor to have a L3 software lookup. Note that this doesn't mean a packet is never sent to the MSFC (punted) but this is the exception rather than the norm.

the The Route Switch Module or RSM seems to be the module that actually runs the MLSP to move MLS-RP entries to MLS-SE entries.

So MLS-RP entries are they in hardware in ASICs or is that software based?

The MLS-SE entries are stored on the switch in a cache and packets are hardware switch by the MLS-SE.

Jon

Joshua Davis · ‎04-15-2010

Ok after spending nearly an hour reading the online cirriculum they never even mention MLS SE.

We are only informed that the route processor is a layer 3 engine that builds the FIB and adjacency tables in software as part of the "control plane" and then relays this information to the interface module a layer 2 engine for packet forwarding in hardware as part of the "data plane"

So I still have no idea how MLS SE fits into Multi Layer Switching. First of all does CEF use MLS SE? If so for what? Would it be used under the Centralized switching model or the Distributed Switching model or both? Is it part of the control plane or data plane? What is its relation to the FIB and adajency tables since those are created by the Route Processor.

Also the Route Processor is supposed to handle access lists and QoS, quality of service and allow switches to manage these items at wire speed, but how does this information get relayed to hardware as the Route Processor is operating in software as part of the layer 3 engine? Does the Route Processor have a layer 2 hardware aspect that the cirriculum is not discussing?

Thank you very much in advance for your patience in helping me understand this issue. The cirriculum and the BCMSN offical guide doesn't really explain this matter in a way that can give me a full understanding and I don't really know where else I could get a basic level understanding of this technology.

Joshua

Joshua Davis · ‎04-16-2010

I found what I was looking for. One document that explained everything. Including MLS SE which indeed is a NFFC (NetFlow Feature Card) module which would place it in 1st generation Route Caching and not Topology-Based Switching.

FYI it is the TCAM table that actually performs the implementation of ACLs and QoS at wire speeds not MLS RP. MLS RP as part of the Control Plane simply passes instructions onto the TCAM table operating in the data plane which actually performs the functions of ACL filtering at wire speets.

http://www.cabrillo.edu/~rgraziani/courses/cis187.html

PowerPoint Presentations
Inter-VLAN Routing, Multilayer Switching and CEF

http://valiente.cabrillo.edu/curriculum/graziani/cis187/presentations/cis187-4-MLS-CEF.ppt

I cannot provide the username password for this but it states clearly on the web page:

Non-Cabrillo Instructors/Students: If you would like access to my materials please email me for the username and password.

Joshua