To improving network performance, switches must address three issues:

∙They must stop unneeded traffic from crossing network segments.

∙They must allow multiple communication paths between segments.

∙They cannot introduce performance degradation.

Routers are also used to improve performance. Routers are typically attached to switches to connect multiple LAN segments. A switch forwards the traffic to the port on the switch to which the destination device is connected, which in turn reduces the traffic to the other devices on the network. Information from the sending device is routed directly to the receiving device. No device other than the router, switch, and end nodes sees or processes the information.

The network now becomes less saturated, more secure, and more efficient at processing information, and precious processor time is freed on the local devices. Routers today are typically placed at the edge of the network and are used to connect WANs, filter traffic, and provide security. See Figure 1.3.

Figure 1.3: Routers and switches

Like bridges, switches perform at OSI Layer 2 by examining the packets and building a forwarding table based on what they hear. Switches differ from bridges by helping to meet the following needs for network designers and administrators:

∙Provide deterministic paths

∙Relieve network bottlenecks

∙Provide deterministic failover for redundancy

∙Allow scalable network growth

∙Provide fast convergence

∙Act as a means to centralize applications and servers

∙Have the capacity to reduce latency

Network Design

When designing or upgrading your network, you need to keep some basic rules of segmenting in mind. You segment your network primarily to relieve network congestion and route data as quickly and efficiently as possible. Segmentation is often necessary to satisfy the bandwidth requirements of a new application or type of information that the network needs to support. Other times, it may be needed due to the increased traffic on the segment or subnet. You should also plan for increased levels of network usage or unplanned increases in network population.

Some areas you need to consider are the types of nodes, user groups, security needs, population of the network, applications used, and the network needs for all the interfaces on the network. When designing your network, you should create it in a hierarchical manner. Doing so provides you with the ability to easily make additions to your network. Another important consideration should be how your data flows through the network.

For example, let’s say your users are intermingled with your servers in the same geographical location. If you create a switched network in which the users’ data must be switched through a number of links to another geographical area and then back again to create a connection between the users and file servers, you have not

14

designed the most efficient path to the destination.

Single points of failure need to be analyzed, as well. As we stated earlier, every large−network user has suffered through his or her share of network outages and downtime. By analyzing all the possible points of failure, you can implement redundancy in the network and avoid many network outages. Redundancy is the addition of an alternate path through the network. In the event of a network failure, the alternate paths can be used to continue forwarding data throughout the network.

The last principle that you should consider when designing your network is the behavior of the different protocols. The actual switching point for data does not have to be the physical wire level. Your data can be rerouted at the Data Link and Network layers, as well. Some protocols introduce more network traffic than others. Those operating at Layer 2 can be encapsulated or tagged to create a Layer−3−like environment. This environment allows the implementation of switching, and thereby provides security, protocol priority, and Quality of Service (QoS) features through the use of Application−Specific Integrated Circuits (ASICs) instead of the CPU on the switch. ASICs are much faster than CPUs. ASICs are silicon chips that provide only one or two specific tasks faster than a CPU. Because they process data in silicon and are assigned to a certain task, less processing time is needed, and data is forwarded with less latency and more efficiency to the end destinations.

In order to understand how switches work, we need to understand how collision domains and broadcast domains differ.

Collision Domains

A switch can be considered a high−speed multiport bridge that allows almost maximum wire−speed transfers. Dividing the local geographical network into smaller segments reduces the number of interfaces in each segment. Doing so will increase the amount of bandwidth available to all the interfaces. Each smaller segment is considered a collision domain.

In the case of switching, each port on the switch is its own collision domain. The most optimal switching configuration places only one interface on each port of a switch, making the collision domain two nodes: the switch port interface and the interface of the end machine.

Let’s look at a small collision domain consisting of two PCs and a server, shown in Figure 1.4. Notice that if both PCs in the network transmit data at the same time, the data will collide in the network because all three computers are in their own collision domain. If each PC and server was on its own port on the switch, each would be in its own collision domain.

Figure 1.4: A small collision domain consisting of two PCs sending data simultaneously to a server.

Switch ports are assigned to virtual LANs (VLANs) to segment the network into smaller broadcast domains. If you are using a node attached to a switch port assigned to a VLAN, broadcasts will only be received from members of your assigned VLAN. When the switch is set up and each port is assigned to a VLAN, a broadcast sent in VLAN 1 is seen by those ports assigned to VLAN 1 even if they are on other switches attached by trunk links. A switch port can be a member of only one VLAN and requires a Layer 3 device such as an internal route processor or router to route data from one VLAN to another.

15

Although the nodes on each port are in their own collision domain, the broadcast domain consists of all of the ports assigned to a particular VLAN. Therefore, when a broadcast is sent from a node in VLAN 1, all the devices attached to ports assigned to VLAN 1 will receive that broadcast. The switch segments the users connected to other ports, thereby preventing data collisions. For this reason, when traffic remains local to each segment or workgroup, each user has more bandwidth available than if all the nodes are in one segment.

On a physical link between the port on the switch and a workstation in a VLAN with very few nodes, data can be sent at almost 100 percent of the physical wire speed. The reason? Virtually no data collisions. If the VLAN contains many nodes, the broadcast domain is larger and more broadcasts must be processed by all ports on the switch belonging to each VLAN. The number of ports assigned to a VLAN make up the broadcast domain, which is discussed in the following section.

Broadcast Domains

In switched environments, broadcast domains consist of all the ports or collision domains belonging to a VLAN. In a flat network topology, your collision domain and your broadcast domain are all the interfaces in your segment or subnet. If no devices (such as a switch or a router) divide your network, you have only one broadcast domain. On some switches, the number of broadcast domains or VLANs that can be configured is almost limitless. VLANs allow a switch to divide the network segment into multiple broadcast domains. Each port becomes its own collision domain. Figure 1.5 shows an example of a properly switched network.

Figure 1.5: An example of a properly switched network.

Note Switching technology complements routing technology, and each has its place in the network. The value of routing technology is most noticeable when you get to larger networks that utilize WAN solutions in the network environment.

Why Upgrade to Switches?

As an administrator, you may not realize when it is time to convert your company to a switched network and implement VLANs. You may also not be aware of the benefits that can occur from replacing your Layer 2 hubs and bridges with switches, or how the addition of some modules in your switches to implement routing and filtering ability can help improve your network’s performance.

When your flat topology network starts to slow down due to traffic, collisions, and other bottlenecks, you may want to investigate the problems. Your first reaction is to find out what types of data are flowing through your network. If you are in command of the network sniffer or other such device, you may begin to find over−utilization errors on the sniffer occurring when the Ethernet network utilization reaches above only 40 percent.

Why would this happen at such a low utilization percentage on the network? Peak efficiency on a flat topology Ethernet network is about 40 percent utilization. Sustained utilization above this level is a strong indicator that you may want to upgrade the physical network into a switched environment.

When you start to notice that your state−of−the−art Pentiums are performing poorly, many network administrators don’t realize the situation may be due to the hundreds of other computers on their flat hub and

16

bridged networks. To resolve the issue, your network administrator may even upgrade your PC to a faster CPU or more RAM. This allows your PC to generate more input/output (I/O), increasing the saturation on the network. In this type of environment, every data packet is sent to every machine, and each station has to process every frame on the network.

The processors in the PCs handle this task, taking away from the processing power needed for other tasks. Every day, I visit users and networks with this problem. When I upgrade them to a switched network, it is typically a weekend job. The users leave on Friday with their high−powered Pentiums stacked with RAM acting like 486s. When they come back Monday morning, we hear that their computers boot up quickly and run faster, and that Internet pages come up instantly.

In many cases, slow Internet access times were blamed on the users’ WAN connections. The whole time, the problem wasn’t their WAN connections—it was their LAN saturated to a grinding halt with frames from every interface on the network.

When network performance gets this bad, it’s time to call in a Cisco consultant or learn how to implement switching. Either way, you are reading this book because you are very interested in switching or in becoming Cisco certified. Consider yourself a network hero of this generation in training.

To fix the immediate problems on your 10BaseT network with Category 3 or Category 4 cabling, you might need to upgrade to Category 5 cabling and implement a Fast Ethernet network. Then you need to ask yourself, is this only a temporary solution for my network? What types of new technologies are we considering? Are we going to upgrade to Windows 2000? Will we be using Web services or implementing Voice Over IP? Do we have any requirements for using multicast, unicast, video conferencing, or CAD applications? The list of questions goes on. Primarily, you need to ask yourself if this is a temporary solution or one that will stand the test of time.

Unshielded Twisted−Pair Cable

Category 3 unshielded twisted−pair (UTP) is cable certified for bandwidths of up to 10Mbps with signaling rates of up to 16MHz. Category 4 UTP cable is cable certified for bandwidths of up to 16Mbps with signaling rates up to 20MHz. Category 4 cable is classified as voice and data grade cabling. Category 5 cabling is cable certified for bandwidths of up to 100Mbps and signaling rates of up to 100MHz. New cabling standards for Category 5e and Category 6 cable support bandwidths of up to 1Gbps.

In many cases, network administrators don’t realize that implementing a switched network will allow your network to run at almost wire speed. Upgrading the backbone (not the wiring), eliminating the data collisions, making the network segments smaller, and getting those users off hubs and bridges is the answer. In terms of per−port costs, this is usually a much cheaper solution. It’s also a solution you can grow with. Of course, a 100Mbps network never hurts; but even a switched 10BaseT network that has been correctly implemented can have almost the same effect of providing your network with increased performance.

Network performance is usually measured by throughput. Throughput is the overall amount of data traffic that can be carried by the physical lines through the network. It is measured by the maximum amount of data that can pass through any point in your network without suffering packet loss or collisions.

Packet loss is the total number of packets transmitted at the speed of the physical wire minus the number that arrive correctly at their destination. When you have a large percentage of packet losses, your network is functioning less efficiently than it would if the multiple collisions of the transmitted data were eliminated.

The forwarding rate is another consideration in network throughput. The forwarding rate is the number of packets per second that can be transmitted on the physical wire. For example, if you are sending 64−byte packets on a 10BaseT Ethernet network, you can transmit a maximum of about 14,880 packets per second.

17

Poorly designed and implemented switched networks can have awful effects. Let’s take a look at the effects of a flat area topology and how we can design, modify, and upgrade Ethernet networks to perform as efficiently as possible.

Properly Switched Networks

Properly switched networks use the Cisco hierarchical switching model to place switches in the proper location in the network and apply the most efficient functions to each. In the model you will find switches in three layers:

∙Access layer

∙Distribution layer

∙Core layer

Note Chapter 2 will introduce the layers at which each switch can be found and the basic configuration steps for both of the command line interfaces.

The Access layer’s primary function is to connect to the end−user’s interface. It routes traffic between ports and broadcasts collision domain traffic to its membership broadcast domain. It is the access point into the network for the end users. It can utilize lower−end switches such as the Catalyst 1900, 2800, 2900, 3500, 4000, and 5000 series switches.

The Access layer switch blocks meet at the Distribution layer. It uses medium−end switches with a little more processing power and stronger ASICs. The function of this layer is to apply filters, queuing, security, and routing in some networks. It is the main processor of frames and packets flowing through the network. Switches found at this layer belong to the 5500, 6000, and 6500 series.

The Core layer’s only function is to route data between segments and switch blocks as quickly as possible. No filtering or queuing functions should be applied at this layer. The highest−end Cisco Catalyst switches are typically found at this layer, such as the 5500, 6500, 8500, 8600 GSR, and 12000 GSR series switches.

How you configure your broadcast and collision domains—whether in a switched network or a flat network topology—can have quite an impact on the efficiency of your network. Let’s take a look at how utilization is measured and the different effects bandwidth can have on different media types and networks.

Network Utilization

Network administrators vary on the utilization percentage values for normal usage of the network. Table 1.1 shows the average utilization that should be seen on the physical wire. Going above these averages of network utilization on the physical wire is a sign that a problem exists in the network, that you need to make changes to the network configuration, or that you need to upgrade the network.

Table 1.1: The average limits in terms of physical wire utilization. Exceeding these values indicates a network problem.

You can use a network monitor such as a sniffer to monitor your utilization and the type of traffic flowing through your network. Devices such as WAN probes let you monitor the traffic on the WAN.

18