Campus infrastructure best practices are aimed at designing highly available modular mechanisms that make networks easier to scale, easier to comprehend, and easier to troubleshoot. This is usually accomplished by delivering deterministic traffic patterns.
Collapsing the Core Layer
The process of collapsing the Core Layer implies combining it with the Distribution Layer and this is one of the first decisions small and medium Enterprises make when analyzing the Enterprise Campus design.
The Core Layer is the aggregation point for all of the other layers and modules in the Enterprise Campus architecture. The main technology features required inside the Core Layer include a high level of redundancy, high-speed switching, and high reliability.
Core Layer devices should be able to implement scalable protocols and technologies and they should support advanced load balancing mechanisms. Typically, the Core Layer infrastructure is based on Layer 3 high-speed switching and uses hardware-accelerated network services. Not all Enterprise Campus environments need a dedicated Core Layer, so in some situations it may be combined with the Distribution Layer to form a collapsed Core Layer architecture, as shown in Figure 2.1 below:
Figure 2.1 – Collapsed Core Layer Architecture
Another frequently used scenario is the one in which an organization has multiple buildings or areas that are composed of an Access Layer and a Distribution Layer. Instead of aggregating the connections through a Core Layer, the Distribution Layer switches are connected in a full-mesh topology. In this full-mesh collapsed Core Layer scenario, a number of connections are needed (n*(n-1), where “n” equals the number of Distribution Layer devices), as illustrated in Figure 2.2 below:

Figure 2.2 – Full-Mesh Distribution Layer
| Note: All of the Distribution Layer (collapsed Core Layer) devices in a single area should have at least dual connections to achieve high availability. |
Although some organizations can choose to use a collapsed Core Layer architecture, implementing a dedicated Core Layer has several advantages:
- The Distribution Layer switches can be connected hierarchically to the backbone switches in the Core Layer
- Less physical cabling
- Less complex routing environment
- Improved scalability
- Easier data and voice integration
- Easier to converge LAN and WAN
- Improved fault isolation
- Improved summarization capabilities
- Allows for identifiable boundaries
Using a collapsed Core Layer and full-mesh architecture requires increasing port density because each Distribution Layer switch would need to connect to other Distribution Layer switches. Alternatives to this issue include having a Layer 2 Core Layer with discrete VLANs on each Core Layer switch, so only two ports per Distribution Layer switch are needed. However this option is not very common in modern networks because of the use of multilayer switches.
Another downside of collapsing the Core Layer is the complexity of adding another network building or area. The new Distribution Layer switches would have to be connected to the full-mesh topology, so the total number of links would increase greatly. This problem can be avoided by using a dedicated Core Layer.
Considering all of the facts presented, large Enterprises that use a collapsed Core Layer model should upgrade to a three-layer network model with a dedicated Core Layer as soon as the appropriate expertise and resources are available. Cisco recommends having redundant connections between the Access Layer switches and the Distribution Layer switches, and between the Distribution Layer switches and the Core Layer switches.
Cisco Non-Stop Forwarding with Stateful Switchover
Cisco Non-Stop Forwarding (NSF) with Stateful Switchover (SSO) is another high availability feature (see Figure 2.3 below), and this can be implemented as a supervisor redundancy mechanism (within a Catalyst 6500 switch). One of the main objectives when implementing this feature is avoiding the interruption of Transmission Control Protocol (TCP) sessions (e.g., Border Gateway Protocol (BGP) neighbor relationships). The integrity of these sessions should be maintained to provide seamless convergence. In Layer 2 environments, SSO can be used on its own, while NSF with SSO can be used in Layer 3 environments.
Figure 2.3 – Cisco NSF with SSO
When this feature is activated, the standby Route Processor (RP) will take control of the router after a hardware or software fault on the active RP. Cisco NSF will continue to forward packets until route convergence is complete, and SSO allows the standby RP to take immediate control and maintain connectivity protocols (e.g., BGP sessions).
Cisco NSF is supported by a wide variety of routing protocols, including the following:
- EIGRP
- OSPF
- IS-IS
- BGP
A router or a multilayer switch that runs one of these protocols can detect internal switchover and take the proper actions to continue forwarding the network traffic, using and leveraging Cisco Express Forwarding (CEF) while recovering route information from other devices, so forwarding continues and no link flapping occurs.
Figure 2.4 – Cisco NSF Recommendations
Analyzing Figure 2.4 above, Cisco provides some recommendations for each layer in the Service Provider Edge module and the Campus Enterprise module, as follows:
- Service Provider Core Layer: The Cisco NSF with SSO feature may provide some benefits, but usually it is not required.
- Service Provider Distribution Layer: This is a good position for NSF-aware routers.
- Service Provider Access Layer: This is the primary deployment position for Cisco NSF with SSO-capable routers.
- Campus Enterprise Access Layer: This is the secondary deployment position for Cisco NSF with SSO-capable routers.
- Campus Enterprise Distribution Layer: This is a good position for NSF-aware routers.
- Campus Enterprise Core Layer: SSO may provide some benefits.
Cisco IOS Software Modularity
To add extra functionality to Cisco 6500 IOS non-modular software (e.g., to patch a router), the new image must be loaded on the active and backup supervisors and a reload must occur before the new software patch can take effect. With Cisco IOS software modularity, this is no longer necessary. The most important advantages Cisco IOS software provides are as follows:
- It minimizes unplanned downtime.
- It makes software modifications much easier to accomplish.
- It can be integrated with the change control policy.
- It allows for process-level policy control.
To fully understand the Cisco IOS software modularity features, it is necessary to understand the different services that can be implemented and its different distributed subsystems, as follows:
- The control plane: Routing and multicast protocols.
- The data plane: ACLs, CEF FIB (Forwarding Information Base), or QoS.
- The management plane: CLI, SNMP, NetFlow, or SSH (Secure Shell).
A Cisco Catalyst 6500 switch that has IOS software modularity features allows several IOS control subsystems to operate in independent processes. To achieve process-level automated policy control, the Cisco Embedded Event Manager (EEM) can be integrated. This allows the offload of time-consuming processes to the network and the acceleration of resolutions for troubleshooting problems. EEM is a combination of processes and its goal is to allow critical system parameters monitoring, such as CPU utilization, SNMP alarms, Syslog events, and interface counters. The key benefits of the IOS software modularity feature include the following:
- Operational consistency: New CLI commands are added.
- Memory protection: Cisco IOS software modularity offers a memory architecture where subprocesses use protected address space only. (Every process and its associated subsystem has its own individual memory space, so memory corruption across the boundaries of the processes are avoided.)
- Fault containment: This increases availability, as issues that occur within one process do not affect other parts of the system.
- Process restartability: The modular processes can be restarted individually (using the “process restart” command).
- Process modularization: Each process is modularized.
- Subsystem ISSUs (In-Service Software Upgrades): Selective system maintenance during runtime is available through individual patches (i.e., versioning and patch management functionality, where patches can be installed and activated without restarting the system).
Spanning Tree Protocol Best Practices
The Spanning Tree Protocol (STP), defined by IEEE 802.1d, is a loop-prevention protocol that allows switches to communicate with each other to discover physical loops in a network. If a loop is found, STP specifies an algorithm that switches can use to create a loop-free logical topology. This algorithm creates a tree structure of loop-free leaves and branches that spans across the Layer 2 topology.
Implementing STP is another key issue when considering Layer 2 Campus infrastructure best practices. Although many designers disable STP at the network edge, Cisco does not recommend this because the loss of connectivity without STP is a much bigger issue than any information that might be revealed through STP to the outside world. The most important issues a network designer should take into consideration when working with STP are when to implement STP and which version of STP should be used.
There are several recommendations for STP implementation. First, if a VLAN spans multiple Access Layer switches to support business applications, STP might be necessary prevent loops. Second, STP could be implemented to support data center applications on server farms or to protect against user-side loops, which can occur in multiple ways:
- Workstation misconfiguration
- Malicious users
- Wiring misconfiguration
When considering the appropriate STP version, Cisco recommends Rapid Per VLAN Spanning Tree+ (RPVST+). In addition, network designers should take advantage of the Cisco STP toolkit, which includes the following features:
- PortFast: This is used to configure a LAN interface as an access port to enter the forwarding state immediately, bypassing the listening and learning states. PortFast should be used when connecting a single end-station to a Layer 2 access port or when dealing with trunks to servers in a data center.
- UplinkFast: This feature offers a 3- to 5-second convergence after a direct link failure and allows load balancing between redundant Layer 2 links using “uplink groups.”
- BackboneFast: This will reduce convergence time and will be activated when a Root port or a blocked port on a network device receives inferior Bridge Protocol Data Units (BPDUs) from its designated switch.
- Loop Guard: This feature moves a non-designated port into the STP loop-inconsistent blocking state if it receives a BPDU. This prevents the port from assuming the designated role and forwarding packets, thus creating a loop.
- Root Guard: This is used to enforce the Root Bridge placement in the network by ensuring the port on which Root Guard is enabled is the designated port. If a switch receives a superior BPDU on a Root-Guard protected port, that specific port is placed into a root-inconsistent state.
- BPDU Guard: When BPDU Guard is enabled on a non-trunking port, it places it into an errdisabled state if it receives a BPDU. When enabling BPDU Guard globally, it shuts down ports in the operational PortFast state that receive BPDUs.
- BPDU Filter: BPDU filtering is used to avoid transmitting BPDUs on a PortFast-enabled port that connects to end systems.
- UDLD: UniDirectional Link Detection (UDLD) permits devices to monitor the physical cable and act when a unidirectional link exists by shutting down the affected LAN port. It is often configured on ports that are linking switches together. UDLD is most often implemented on optical fiber links, but it also supports copper Ethernet technology. Cisco recommends enabling UDLD in the “aggressive” mode, instead of the “normal” mode, in all environments that have optical fiber links because it actively tries to re-establish the state of the port once the connection is down.
Figure 2.5 – Cisco STP Features
| Note: All of the STP features in Figure 2.5 above are covered in detail in the CCNP Switch exam study materials. |
Trunking Best Practices
Some of the most important recommendations when using trunking include the following:
- Use 802.1Q instead of Inter-Switch Link (ISL) for encapsulation, because in this case, the VLAN tag is internal to the Ethernet frame and the frame size is not affected.
- Set the native VLAN to an unused VLAN.
- Place all of the unused ports in a non-functional VLAN to increase network security.
- Use VLAN Trunk Protocol (VTP) “transparent” mode.
- Set Dynamic Trunking Protocol (DTP) to “desirable” mode with encapsulation “negotiate” when configuring switch-to-switch links that will carry multiple VLANs.
- Manually prune unused VLANs from the trunk interface to avoid the propagation of broadcasts.
- Hard code all of the access ports (disable trunks on host ports) as a security countermeasure to VLAN hopping.
EtherChannel Recommendations
A very common technology used to achieve media redundancy and increase bandwidth is EtherChannel. This is a Layer 2 or Layer 3 logical bundling or channel aggregation technique that can be used between switches and other devices (usually other switches). The bundled links appear as one logical link on the specific devices, as shown in Figure 2.6 below:
Figure 2.6 – Example of an EtherChannel
EtherChannels are important when using STP because when all of the physical links look like one logical link, STP will not consider them a possible loop threat and will not shut down any link from that specific bundle. Therefore, the links in an EtherChannel can be dynamically load balanced, without interference from STP. Cisco switches support two implementation options (protocols) when using EtherChannels:
- Port Aggregation Protocol (PagP) – a Cisco proprietary protocol
- Link Aggregation Control Protocol (LACP) – an open standard protocol
EtherChannel is a very popular technology and is typically deployed between the Distribution Layer and the Core Layer or between Core Layer devices where increased availability and scalability is needed. However, port aggregation is usually disabled on interfaces that are facing end-users. The recommended configuration when using PagP for Layer 2 EtherChannels is to set the port modes to “desirable” on all of the members of the bundle to ensure that individual link failures will not lead to an STP failure. The recommended practice for LACP is to configure both sides of the link in the “active” mode.




