Campus Design Terminology
This section explains some of the terminology that is commonly used to describe network designs. The discussion begins with a review of the Intermediate Distribution Frame/Main Distribution Frame (IDF/MDF) terminology that has been borrowed from the telephone industry. It then looks at a three-level paradigm that can be very useful.
For years, the telephone industry has used the terms Intermediate Distribution Frame (IDF) and Main Distribution Frame (MDF) to refer to various elements of structured cabling. As structured cabling has grown in popularity within data-communication circles, this IDF/MDF terminology has also become common.
The following sections discuss some of the unique requirements of switches placed in IDF and MDF closets. In addition to these specialized requirements, some features should be shared across all of the switches. For new installations, all of the switches should offer a wide variety of media types that include the various Ethernet speeds and ATM. FDDI and Token Ring support can be important when migrating older networks. Also, because modern switched campus infrastructures are too complex for the “plug-it-in-and-forget-it” approach, comprehensive management capabilities are a must.
IDF wiring closets are used to connect end-station devices such as PCs and terminals to the network. This “horizontal wiring” connects to wall-plate jacks at one end and typically consists of unshielded twisted-pair (UTP) cabling that forms a star pattern back to the IDF wiring closet. As shown in Figure 14-1, each floor of a building generally contains one or more IDF switches. Each end station connects back to the nearest IDF wiring closet. All of the IDFs in a building generally connect back to a pair of MDF devices often located in the building’s basement or ground floor.
Figure 14-1. Multiple IDF Wiring Closets
Given the role that they perform, IDF wiring closets have several specific requirements:
- Port density— Because large numbers of end stations need to connect to each IDF, high port density is a must.
- Cost per port— Given the high port density found in the typical IDF, cost per port must be reasonable.
- Redundancy— Because several hundred devices often connect back to each IDF device, a single IDF failure can create a significant outage.
- Reliability— This point is obviously related to the previous point, however, it highlights the fact that an IDF device is usually an end station’s only link to the rest of the world.
- Ease of management— The high number of connections requires that per-port administration be kept to a minimum.
Because of the numerous directly connected end users, redundancy and reliability are critical to the IDF’s role. As a result, IDFs should not only utilize redundant hardware such as dual Supervisors and power supplies, they should have multiple links to MDF devices. Fast failover of these redundant components is also critical.
IDF reliability brings up an interesting point about end-station connections. Outside of limited environments such as financial trading floors, it is generally not cost-effect to have end stations connected to more than one IDF device. Therefore, the horizontal cabling serves as a single point of failure for most networks. However, note that these failures generally affect only one end station. This is several orders of magnitude less disruptive than losing an entire switch. For important end stations such as servers, dual-port network interface cards (NICs) can be utilized with multiple links to redundant server farm switches.
The traditional device for use in IDF wiring closets is a hub. Because most hubs are fairly simple devices, the price per port can be very attractive. However, the shared nature of hubs obviously provides less available bandwidth. On the other hand, routers and Layer 3 switches can provide extremely intelligent bandwidth sharing decisions. On the downside, these devices can be very expensive and generally have limited port densities.
To strike a balance between cost, available bandwidth, and port densities, almost all recently deployed campus networks use Layer 2 switches in the IDF. This can be a very cost-effective way to provide 500 or more end stations with high-speed access into the campus backbone.
However, this is not to say that some Layer 3 technologies are not appropriate for the wiring closet. Cisco has introduced several IDF-oriented features that use the Layer 3 and 4 capabilities of the NetFlow Feature Card (NFFC). As discussed in Chapter 5, “VLANs,” and Chapter 11, “Layer 3 Switching,” Protocol Filtering can be an effective way to limit the impact of broadcasts on end stations. By allowing a port to only output broadcasts for the Layer 3 protocols that are actually in use, valuable CPU cycles can be saved. For example, a broadcast-efficient TCP/IP node in VLAN 2 can be spared from being burdened with IPX SAP updates. IGMP Snooping is another feature that utilizes the NFFC to inspect Layer 3 information. By allowing the Catalyst to prune ports from receiving certain multicast addresses, this feature can save significant bandwidth in networks that make extensive use of multicast applications. Finally, the NFFC can be used to classify traffic for Quality of Service/Class of Service (QoS/COS) purposes.
The most important IDF concerns are cost, port densities, and redundancy.
IDF devices collapse back to one or more Main Distribution Frame (MDF) devices in a star-like fashion. Each IDF usually connects to two different MDF devices to provide adequate redundancy. Some organizations place both MDF devices in the same physical closet and rely on disparate routing of the vertical cabling for redundancy. Other organizations prefer to place the MDF devices in separate closets altogether. The relationship between buildings and MDFs is not a hard rule—larger buildings might have more than two MDF switches, whereas a pair of redundant MDF devices might be able to carry multiple buildings that are smaller in size.
Figure 14-2 shows three buildings with MDF closets. To meet redundancy requirements, each building generally houses two MDF devices. The MDF devices can also be used to interconnect the three buildings (other designs are discussed later).
Figure 14-2. MDF Closets
MDF closets have a different set of requirements and concerns than IDF closets:
- High availability
- Routing capabilities
Given that they act as concentration points for IDF traffic, MDF devices must be able to carry extremely high levels of traffic. In the case of a Layer 2 switch, this bandwidth is inexpensive and readily available. However, as is discussed later in this chapter, many of the strategies to achieve robust and scalable designs require routing in the MDF. Achieving this level of Layer 3 performance can require some careful planning. For more information on Layer 3 switching, see Chapter 11. Issues associated with Layer 3 switching are also addressed later in this chapter and in Chapter 15.
High availability is an important requirement for MDF devices. Although the failure of either an MDF or IDF switch potentially affects many users, there is a substantial distinction between these two situations. As discussed in the previous section, the failure of an IDF device completely disables the several hundred attached end stations. On the other hand, because MDFs are almost always deployed in pairs, failures rarely result in a complete loss of connectivity. However, this is not to say that MDF failures are inconsequential. To the contrary, MDF failures often affect thousands of users, many more than with an IDF failure. This requires as many features as possible that transparently reroute traffic around MDF problems.
In addition to the raw Layer 3 performance discussed earlier, other routing features can be important in MDF situations. For example, the issue of what Layer 3 protocols the router handles can be important (IP, IPX, AppleTalk, and so forth). Routing protocol support (OSPF, RIP, EIGRP, IS-IS, and so on) can also be a factor. Support for features such as DHCP relay and HSRP can be critical.
Three types of devices can be utilized in MDF closets:
- Layer 2 switches
- Hybrid, “routing switches” such as MLS
- “Switching routers” such as the Catalyst 8500
The first of these is also the simplest—a Layer 2 switch. The moderate cost and high throughput of these devices can make them very attractive options. Examples of these devices include current Catalyst 4000 models and traditional Catalyst 5000 switches without a Route Switch Module (RSM) or NFFC.
However, as mentioned earlier, there are compelling reasons to use Layer 3 processing in the MDF. This leads many network designs to utilize the third option, a Layer 3 switch that is functioning as a hardware-based router, what Chapter 11 referred to as a switching router. The Catalyst 8500 is an excellent example of this sort of device.
Cisco also offers another approach, Multilayer Switching (MLS), that lies between the previous two. MLS is a hybrid approach that allows the Layer 2-oriented Supervisors to cache Layer 3 information. It allows Catalysts to operate under the routing switch form of Layer 3 switching discussed in Chapter 11. A Catalyst 5000 with an RSM and NFFC is an example of an MLS switch. Other examples include the Catalyst 5000 Route Switch Feature Card (RSFC) and the Catalyst 6000 Multilayer Switch Feature Card (MSFC).
It is important to understand the differences between the routing switch (MLS) and switching router (Catalyst 8500) styles of Layer 3 switching. These concepts are discussed in detail in Chapter 11.
Although the switching router (8500) and routing switch (MLS) options both offer very high throughput at Layer 3 and/or 4, there are important differences. For a thorough discussion of the technical differences, please see Chapter 11. This chapter and Chapter 15 focus on the important design implications of these differences.
The most important MDF factors are availability and Layer 3 throughput and capabilities.
Three-Layer Campus Network Model: Access, Distribution, Core
The IDF/MDF terminology discussed in the previous section describes the world in terms of two layers. However, MDF interconnections can often be better described with a third layer. For this reason, it is often useful to describe campus (and WAN) networks in terms of a three-layer model that more accurately describes the unique requirements of the inter-MDF connections. Geoff Haviland’s excellent Cisco Internetwork Design (CID) course has popularized the use of the terms access, distribution, and core to describe these three layers. Figure 14-3 illustrates the three-layer model.
Figure 14-3. The Three-Layer Design Model
Each of these layers is briefly discussed in the following three sections.
The IDF closets are termed access layer closets under the three-layer model. The idea is that the devices deployed in these closets should be optimized for end-user access. Access layer requirements here are the same as those discussed in the IDF section: port density, cost, resiliency, and ease of management.
Under the three-layer model, MDF devices become distribution layer devices. The requirement for high Layer 3 throughput and functionality is especially important here.
In campus networks, the term access layer is synonymous with IDF, and distribution layer is equivalent to MDF.
The connections between the MDF switches become the core layer under the three-layer model. As is discussed in detail later, some networks have a very simple core consisting of several inter-MDF links or a pair of Layer 2 switches. In other cases, the size of the network might require Layer 3 switching within the core. Many networks utilize an Ethernet-based core; others might use ATM technology.
In general, the terms access layer and distribution layer are used interchangeably with IDF and MDF. However, the IDF/MDF terms are used most often when discussing two-layer network designs; the access/distribution/core terminology is used when explaining three-layer topologies.