Campus, as it applies to networking, typically describes the routers, switches, network appliances, and servers that make up the networking infrastructure for a set of buildings located in close proximity. A campus can be the manufacturing site of a large corporation, the headquarters for a bank, or a college campus.
Campus networks are characterized by high-speed connectivity. A design goal for campus networks is to separate buildings, floors, workgroups, and server farms into smaller Layer 3 groups to prevent network faults from affecting large populations of users. Layer 3 routers provide natural boundaries against debilitating network problems such as broadcast storms and loops.
Over time, the hierarchical approach to network design has proven the most effective. The three primary layers of a hierarchical campus follow:
Backbone or core—. The backbone is the central thoroughfare for corporate traffic. All other parts of the network eventually feed into the backbone. You should design the core to switch packets as quickly as possible. This level should not include operations that might slow the switching of the packet: the distribution layer should handle any packet manipulation or filtering that needs to occur.
Distribution—. The distribution layer provides policy-based connectivity and boundaries between the access layers and the core. For example, a building of 20 floors might have a distribution network that connects each of the floors with the backbone. It is at this layer that packets should be filtered or manipulated. Therefore, once packets are “prepped,” the core simply needs to switch them quickly to the destination distribution location.
Access—. The access layer provides user access to the network. It is at this point that users are permitted (or denied) access into the corporate network. Typically, each person sitting at a desk has a cable that runs back to a wiring closet and connects to a switch; hence, this level is where the user “accesses” the network.
The distribution and core layers of the network provide vital services by aggregating groups of users and services. Therefore, if a distribution or core device dies, it can affect large communities of users. For this reason, reducing the chance of failure in these layers reduces and possibly prevents unnecessary and unplanned outages. Redundant network paths, redundant hardware, and fault-tolerant–related network protocols (such as Hot Standby Router Protocol) all aid in the ability of a network to recover quickly (and, you hope, transparently to the users) after a failure.
Companies prefer to reduce the number of routed protocols traversing the backbone or core layer. In the 1990s before the massive popularity of TCP/IP and the Internet, backbones carried the predominant protocols of the day: Novell, DECnet, AppleTalk, NetBIOS, and Banyan VINES. So many protocols complicated design issues. As TCP/IP became the de-facto networking protocol, companies worked to eliminate the non-IP traffic from the backbone.
At-A-Glance—Hierarchical Campus Design
Campus networks represent a enormous investment for businesses. When correctly designed, a campus network can enhance business efficiency and lower operational cost. Additionally, a properly designed network can position a business for future growth.
A modular or hierarchical network consists of building blocks that are easier to replicate, redesign, and grow. Each time you add or remove a module, you don’t need to redesign the whole network. You can put distinct blocks in service and out of service without impacting other blocks or the core of the network. This setup greatly enhances troubleshooting, aides fault isolation, and eases management.
The hierarchy consists of three functional divisions (or layers):
Access layer—. First level of access to the network. Layer 2 switching, security, and quality of service (QoS) happen at this layer.
Distribution layer—. Aggregates wiring closets and provides policy enforcement. When the network uses Layer 3 protocols at this layer, it realizes benefits such as load balancing, fast convergence, and scalability. This layer also provides first-hop default gateway redundancy to end stations.
Core layer—. Backbone of the network. This layer is designed to be fast converging, highly reliable, and stable. Also designed with Layer 3 protocols, the core provides load balancing, fast convergence, and scalability.
At-A-Glance—Campus Design Best Practices
The virtual local-area network (VLAN) organizes physically separate users into the same broadcast domain. The use of VLANs improves performance, security, and flexibility. VLANs also decrease the cost of arranging users because they require no extra cabling.
VLANs should each be on their own subnet. This arrangement is called Layer 2 to Layer 3 VLAN mapping. It allows for ease of route summarization and troubleshooting.
(You should avoid setting up campus-wide VLANs because they can slow convergence.)
</division> </division><division> <title>High Availability</title>High availability refers to the network’s ability to recover from different types of failures. A sound design easily achieves network stability. Troubleshooting is easier, and human error is reduced. You should design high availability at many layers:
Layer 1—. Redundant links and hardware provide alternate physical paths through the network.
Layers 2/3—. Protocols such as Spanning Tree Protocol (STP), Hot Standby Router Protocol (HSRP), and other routing protocols provide alternate path awareness and fast convergence.
Application availability—. The application server and client processes must support failover for maximum availability.
Redundancy is a key part of designing a highly available network. Although some redundancy is good, too much redundancy can actually be bad for a network. Issues with convergence (the network’s ability to recover from a bad link) can result. Too much redundancy makes troubleshooting and management difficult. (This figure shows redundancy gone wrong.)
</division> <division> <title>Oversubscription</title>
Oversubscription occurs when the network has more traffic-generating endpoints than it can accommodate at a single time. Most networks are built with some amount of oversubscription. In the figure, the network has a 20:1 oversubscription rate from access to distribution and a 4:1 oversubscription rate from distribution to core. Networks should use QoS to ensure that they do not drop or delay real-time traffic such as voice and video or critical data such as SAP traffic.
</division> </division>
At-A-Glance—Small and Medium Campus Design
Size does matter. The following sections list some general rules of thumb for various campus sizes.
<division> <title>Small Campus</title>This figure shows the recommended design for a campus with fewer than 200 edge ports. This design has collapsed the distribution and core into a single layer, which limits scaling to a few access switches.
</division> <division> <title>Medium Campus</title>
This design is appropriate for campuses with 200–1000 ports. A separate distribution layer allows for future growth. A redundant core ensures high availability and allows equal-cost paths.
</division> </division>
At-A-Glance—Large Campus Design
This high-performance design suits campuses with more than 1000 ports or those seeking cutting-edge design. A feature-rich core aggregates many distribution switches. You can easily add buildings and server farms anytime.
You must carefully consider services such as QoS, IP multicast, and security in any campus design strategy. Other chapters discuss these topics.
</division>