Overview of VLAN Characteristics

Before diving into the details of performance and fault management for VLANs, we define some of the relevant terms and features of the technology. We'll look specifically at the following:

• Logical ports versus physical ports

• VLAN trunking

• Spanning tree as it applies to VLANs

This is not an exhaustive discussion regarding VLANs, just a brief overview. For more details regarding VLAN architecture, refer to the documentation on CCO or some of the specific references cited at the end of this chapter.

Logical Versus Physical Ports

One of the main reasons for lack of system resources on Catalyst Series switches is the presence of too many logical ports configured or allowed on the switch. Each VLAN instance configured on a trunk port runs its own instance of spanning tree, which can cause extensive CPU resources to be used. Refer to Chapter 11, "Monitoring Network Systems Processes and Resources," for specifics on the switch resources.

A logical port is a summation of all physical ports installed on a switch plus the amount of VLANs configured on all trunk ports, assuming spanning tree is active for all those VLANs see Equation 15-1.

Equation 15-1

where the sum of all logical ports equals:

400 for Supervisor Engine I (with 20-MB DRAM)

1500 for Supervisor Engine II and III F

4000 for Supervisor Engine III

We'll refer to this formula throughout the chapter when discussing logical ports.

NOTE

Equation 15-1 is taken from the Catalyst Series Software release notes. The Usage Guidelines and Restrictions section includes the following comments:

Ensure that the total number of logical ports across all instances of spanning-tree for different VLANs does not exceed the maximum number supported for each supervisor engine type and memory configuration. Use Equation 15-1 to compute the total number of logical ports on the switch.

If you enable numerous memory-intensive features concurrently (such as VTP pruning, VMPS, EtherChannel, and RMON), or if there is switched data traffic on the management VLAN, the maximum number of supported logical ports is reduced.

TIP

Based on past experiences and field trials, the recommended practical maximum number of logical ports is as follows:

• 250 for Supervisor Engine I (with 20 MB DRAM)

• 1000 for Supervisor Engine II

• 2500 for Supervisor Engine III

VLAN Trunking

Trunking in VLAN environments is a way to send multiple VLANs over one physical port using some kind of encapsulation method such as Cisco's Inter-Switched Link (ISL), ATM LANE, IEEE's 802.10, or IEEE 802.1q. By trunking VLANs, you eliminate the need for multiple ports, one per VLAN, to interconnect two switches. Trunk ports typically connect directly to upstream switches, such as core or distribution switches, from closet switches. VLANs that traverse multiple buildings or campuses more than likely require the use of trunk ports somewhere in the network design. But which trunk encapsulation method is used depends on the network infrastructure put in place.

For example, if you have Fast-Ethernet, Category 5 twisted pair and all Cisco devices in the campus, you probably use Cisco's proprietary ISL or Inter-Switched Link VLAN trunking encapsulation. If you have an FDDI ring connecting campuses together, you probably use 802.10 as the VLAN trunking encapsulation. If your network has an ATM core with switches directly attached to the ATM "cloud," you probably use LANE as your VLAN trunking encapsulation. If you are standardizing on Ethernet trunking and trying to get away from proprietary trunking methods, 802.1q encapsulation is probably going to be used. For more information regarding the different VLAN encapsulation methods, refer to the Cisco documentation available on CCO.

NOTE

The newer switches and port architecture, such as the 6500s and Gigabit Ethernet, utilize the 802.1q trunking protocol. This protocol is not discussed much in this chapter. Instead, a more general discussion of trunking is provided to help calculate logical ports on a switch.

Figure 15-1 illustrates how trunks are applied in VLAN environments. The bold links between Switch E and the rest of the switches are identified as trunk ports. Notice the different VLANs 10 and 20 traversing multiple switches. The trunk ports are used to send data from both of these VLANs over one physical port.

Figure 15-1. VLAN Trunking

VLAN trunking comes into play with network management when looking at logical ports versus physical ports or when looking at trunk utilization (refer to Chapters 12 and 4 for interface utilization calculations and explanations).

Spanning Tree (802.1d)

When creating fault-tolerant switched internetworks, a loop-free path must exist between all nodes in the network. A spanning tree algorithm is used to calculate the best loop-free path through a Catalyst-switched network. Spanning tree packets or BPDUs (Bridge Protocol Data Units) are sent and received by switches in the network at regular intervals. These packets are not forwarded by the switches participating in the spanning tree, but are instead processed to determine the spanning tree itself. The IEEE 802.1D bridge protocol, sometimes referred to as Spanning Tree Protocol or STP, processes the BPDUs or spanning tree packets for Catalyst Series switches.

The Catalyst Series switches normally do use STP on all VLANs. The STP detects and breaks loops by placing some connections in a blocked state; blocked connections are activated in the event of a primary connection going down. A separate STP runs within each configured VLAN, ensuring valid Layer 2 topologies throughout the network.

The supported STP states are as follows:

• Disabled

• Forwarding

• Learning

• Listening

• Blocking

The state for each port initially is set by the configuration, either forwarding (if the "portfast" or "backbonefast" feature is enabled) or blocking and later modified by the STP process. After the port state is set, the 802.1D bridge specification (RFC 1493) determines whether the port forwards or blocks packets.

If not properly designed, the spanning tree feature can cause more headaches to network managers than any other issue in the network. At a minimum, therefore, we recommend the following configuration:

• Root bridges and backup root bridges statically defined in the Core or Distribution switches when using spanning tree.

• Spanning tree should be activated on all trunk ports for all VLANs.

• Optionally, you can activate spanning tree on end-user switch ports as well, in case a loop is mistakenly created in the network. By default, STP is enabled on all ports, so enabling it typically is not required.

Such small precautions can prevent your network from grinding to a halt when a loop is created.

Spanning tree affects network management in the context of logical ports versus physical ports, as well as fault management of VLANs. We'll look later in this chapter at MIBs from the BRIDGE-MIB as it pertains to spanning tree as well as the SNMP traps associated with STP.

For more detailed information on designing switched networks, refer to the Design guides on CCO relating to Switched Internetworks.