HARDWARE cef or SOFTWARE cef ?

Hi. I try to understand the L3 switching issues on the Catalyst range switches. As I already know MLS was the mechanism of the L3 switching on the older sort of Catalysts (5000/6500MSFC1). But the MLS has evolved to the CEF switching.

Do CEF mean L3 switching ? And what's difference between CEF on router and on L3 switch ?

What about to MLS and CEF working the same time on e.g. 6500? Are they mutually exclusive or they work together in some manner? In what manner?

When I issue "sh platform forward ... " on 3750, does it mean HARDWARE CEF((I call it)? How about "sh ip cef " .

It is very confusing. I don't know whether CEF equal to L3 switching or there is HARDWARE CEF and SOFTWARE CEF ?

For example, I can issue "sh ip cef 10.1.1.5" on 6509 RP and "sh mls cef 10.1.1.5" on 6509 SP. What's difference between RP's cef and SP's cef ?

Could someone explain the issue?

thanks!

Reply to
comp.dcom.sys.cisco
Loading thread data ...

This is a simplified explanation of the MLS and CEF.

CEF is a method for populating a L3 switching engine. The L3 switching engine does not use the routing table to do lookups because the routing table is kept in regular memory and is too slow for doing lookups, and contains way more information that is needed to L3 switch the packet. CEF is a method that used to program the hardware L3 Asics from the routing table. Before CEF there was "fast-switching", but this was not as efficient as CEF. In the L3 engine, if the destination is not found, the packet is sent to the routing platforms processor for it to deal with (the packet is said to be "process switched", which is also known as "slow path"). With "fast-switching", only previously seen destinations are processed by the L3 engine, so the first packet to a specific destination is process-switched. The processor looks up the destination in the routing table copies the packet to the appropriate destination, and then programs the L3 engine with the destination and outbound interface. CEF on the otherhand, programs the L3 asic with all destinations in the routing table, and updates to the routing table are propagated to the L3 asics. With CEF, if a routing destination exists in routing table, than it is L3 switched, it doesn't matter if its the first packet, or the 10th packet. The only packets process-switched are the ones that need to be because there is no destination available. The use of CEF dramatically increased the performance of all routing platforms, because the biggest limiting factor is the speed of the processor. The fewer that packets that need to be process-switched, the more packets the router can handle.

On a router platform, you have many different types of layer 2 interfaces. Token-Ring, Ethernet, ATM, Serial, etc, so the L3 switching engine takes packets from the input queue on one interface, and "moves" them to the output queue on another interface The layer 2 headers are striped off before going to the L3 switching engine, and the output interface is responsible for adding the appropriate l2 header to the frame before sending it out. (The output queue is where traffic-shaping, queuing, WRED, and other QoS mechanisms are applied.) On a LAN switching platform, all of the interfaces are of the same type, i.e. Ethernet, so there is no need strip off the L2 header. This layer 2 switching is MLS. With MLS, an incoming packet goes into the L2 switching engine, the destination MAC is looked up in the L2 switching ASICs. If it is a known destination MAC, it is "switched" to appropriate destination interface. If the destination MAC is a MAC address of the switch, then the packet is "switched" to the L3 switching engine, and processed appropriately. The L3 switching engine modifies the L2 and L3 headers, and then passes it back to the L2 asics (MLS) and it then switched out the appropriate interface. The L3 switching engine on LAN switch is a very different architecture than that used on a router, because all of the interfaces are of the same type As asics get bigger and more sophisticated, the line between L2 and L3 on a LAN switch continues to get fuzzier and fuzzier. Because the packets are L2 "switched", you don't have the same QoS mechanisms that you have on router. With Layer 2, you do have CoS (class of service), but this is not the same thing as QoS (quality of service), but this is whole other discussion.

In short MLS is Layer 2 switching, and CEF is used to manage the L3 switching engine. On a 6500 in hybrid mode, you can see a clear delineation between the L2 and L3. CatOS is used to manage the L2, and IOS is used to manage the L3. In native mode, a single user interface is presented to the user, but the L2 and L3 functions are still separate entities that are manage behind the covers by IOS.

Scott

"comp.dcom.sys.cisco" wrote in message news: snipped-for-privacy@t31g2000cwb.googlegroups.com...

Reply to
thrill5

thanks!

You said " MLS is Layer 2 switching", but CISCO said that MLS is L3 switching method on 5500 and 6500 with sup1.

I'm confusing.

Reply to
comp.dcom.sys.cisco

Cabling-Design.com Forums website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.