Understanding and Configuring Multilayer Switching
This chapter covers the following topics:
- Multilayer Switching Architecture on Catalyst Switches
- MLS Memory Table Architecture
- CEF-Based MLS Configuration, Verification, and Troubleshooting
The purpose of this chapter is to provide you with details, architecture, and methods of multilayer switching on Catalyst switches. An understanding of multilayer switching is necessary for network designers, administrators, and operators for deployment and troubleshooting purposes.
The term multilayer switching refers to the ability of a Catalyst switch to support switching and routing of packets in hardware, with optional support for Layers 4 through 7 switching in hardware as well. As mentioned in Chapter 1, "Introduction to Building Cisco Multilayer Switched Networks," switching and routing in hardware (hardware switching) yields high throughput at or near line rate even with all ports sending traffic simultaneously.
With multilayer switches, the prime area of focus is often raw performance. Multilayer switches tend to have packet-switching throughputs in the millions of packets per second (pps), while traditional general-purpose routers have evolved from the 100,000 pps range to just over a million pps. Cisco Catalyst switches achieve this high rate of performance by using hardware switching.
For Catalyst switches to perform hardware switching, a route processor (Layer 3 engine) must download software-based routing, switching, access lists, QoS, and other information to the hardware for packet processing. To accomplish multilayer switching (packet processing in hardware), Cisco Catalyst switches use either the traditional multilayer switching (traditional MLS) or the Cisco Express Forwarding (CEF)-based MLS architecture. Traditional MLS is a legacy feature, whereas all leading-edge Catalyst switches support CEF-based multilayer switching (CEF-based MLS). Table 9-1 illustrates Catalyst switch support of traditional MLS and CEF-based MLS.
Table 9-1. Layer 3 Catalyst Switch Support of Traditional MLS and CEF-Based MLS
Catalyst Switch Family
MLS Mode of Operation
Catalyst 3550, 3560, and 3750
Catalyst 4000 or 4500 running Cisco IOS and using Supervisor Engine II+, III, IV, or V
Catalyst 5000 or 5500 with RSM or RSFC and NFFC or NFFC II
Catalyst 6500 with Supervisor Engine I with MSFC
Catalyst 6500 with Supervisor Engine II, Supervisor 32, or Supervisor 720 with MSFC
In terms of CCNP BCMSN exam preparation, you should focus on the following:
- The path of a packet with CEF-based MLS
- Understanding the differences between centralized and distributed switching
- CEF troubleshooting commands
The section dealing with traditional MLS is outside the scope of the current CCNP BCMSN exam.
Understanding Traditional MLS
MLS enables specialized application-specific integrated circuits (ASIC) to perform Layer 2 rewrite operations of routed packets. Layer 2 rewrites include rewriting the source and destination MAC addresses and writing a recalculated cyclic redundancy check (CRC). Because the source and destination MAC addresses change during Layer 3 rewrites, the switch must recalculate the CRC for these new MAC addresses.
For Catalyst switches that support traditional MLS, the switch learns Layer 2 rewrite information from an MLS router via an MLS protocol. Another name for traditional MLS is NetFlow-based switching. With traditional MLS, the Layer 3 engine (route processor) and switching ASICs work together to build Layer 3 entries on the switch. Each of these Layer 3 entries is populated in one of the following ways:
- Source IP address exclusively (S)
- Source and destination IP address (S/D)
- Full Flow Information with Layer 4 protocol information (FFI)
With traditional MLS, the switch forwards the first packet in any flow to the Layer 3 engine for processing using software switching. After the routing of the first packet in the flow, the Layer 3 engine programs the hardware-switching components for routing for subsequent packets. Figure 9-1 illustrates these fundamentals of traditional MLS.
Figure 9-1 Fundamentals of Traditional MLS
In Figure 9-1, when workstation A sends a packet to workstation B, workstation A sends the packet to its default gateway. In this figure, the default gateway is the RSM. The switch (MLS-SE) recognizes this packet as an MLS candidate packet because the destination MAC address matches the MAC address of the MLS router (MLS-RP). As a result, the switch creates a candidate entry for this flow. Next, the router accepts the packets from workstation A, rewrites the Layer 2 destination MAC address and CRC, and forwards the packet to workstation B. The switch refers to the routed packet from the RSM as the enabler packet. The switch, upon seeing both the candidate and enabler packets, creates an MLS entry in hardware such that the switch rewrites and forwards all future packets matching this flow. The MLS switched packet arrow in Figure 9-1 indicates this flow. This behavior is very different from CEF-based MLS.
On the Catalyst 5000 family of switches, MLS requires specialized supervisors and line modules and a router. In addition, MLS on Catalyst 5000 requires manual configuration. On the Catalyst 6500 family of switches with a Supervisor I Engine with an MSFC, all the line modules support MLS. In addition, this hardware combination uses MLS by default, and configuration of MLS is not required. For more details on and examples of the MLS architecture, consult the following technical documents at Cisco.com.
"Configuring and Troubleshooting IP MLS on Catalyst 6000 Switches with an MSFC," Document ID: 10566
"Troubleshooting IP Multilayer Switching," Document ID: 10554
Traditional MLS is a legacy feature; all leading-edge and future Catalyst switches support CEF-based MLS. The next section and the remainder of this chapter discuss CEF-based MLS and its architecture.