LAN Interconnection and Internetworking

LAN Interconnection and Internetworking

The realm of LAN interconnection devices offers a number of options. This section discusses the nature of hubs, LAN switches, virtual LANs (VLANs), bridges, routers, and IP switches.

Hubs

Hubs interconnect the wiring that's connected to workstations. They are a building block of most networks. There are three major types of hubs:

         Active Active hubs regenerate and retransmit signals, just as a repeater does. Because hubs typically have 8 to 12 ports for network computers to connect to, they are sometimes called multiport repeaters. Active hubs require electrical power to run (that's why they're called active).

         Passive Passive hubs serve as connection points and do not regenerate the signal; the signal simply passes through the hub. They do not require electrical power to run. Wiring panels and punchdown blocks are examples of passive hubs.

         Hybrid Hybrid hubs accommodate several different types of cables.

You can connect hubs together to expand a hub network. The advantages of hubs are that they make it easy to change or expand wiring systems, they use different ports to accommodate different cabling types, and they centralize the monitoring of network activity and traffic. Hubs, which are sometimes called concentrators or multistation access units (MSAUs), can also eliminate the need for NICs with onboard transceivers at each node or personal computer.

A group of transceivers can all be located in and managed by an intelligent hub. Intelligent hubs are modular and chassis based, with slots that accommodate the user's choice of interface modules such at Ethernet, Token Ring, or FDDI for connectivity to LANs, WANs, or other network devices. The number of ports on the NIC determines the number of users in the particular star. Intelligent hubs often provide integrated management and internetworking capabilities, as well as Simple Network Management Protocol (SNMP)-based network management. New generations also offer bridging, routing, and switching functions.

Figure 8.7 shows a network that uses a combination of interconnection devices. Intelligent hubs provide connectivity between workstations that comprise a given cluster. An internal backbone is used to internetwork the intelligent hubs to move between different clusters. Those intelligent hubs then connect into a backbone router for purposes of WAN, or campuswide, connectivity.

Figure 8.7. Using interconnection devices

LAN Switches

LAN switches are a very cost-effective solution to the need for increased bandwidth in workgroups. Each port on the switch delivers a dedicated channel to the device or devices attached to that port, thereby increasing the workgroup's total bandwidth and also increasing the bandwidth available to individual users.

Figure 8.8 shows a simple example of a switched Ethernet configuration. One workstation requires 10Mbps on its own, so it has the full services of a 10Mbps port on the switched Ethernet card. Five workstations, on the other hand, each need 2Mbps, so one 10Mbps port serves all five workstations. These five workstations connect into a hub, and that hub connects into the actual port. Servers have extra bandwidth requirements the ones in Figure 8.8 require 25Mbps so they are each served by a bonding of several 10Mbps ports.

Figure 8.8. An example of a switched Ethernet configuration

The key applications for LAN switches are to interconnect the elements of a distributed computing system, to provide high-speed connections to campus backbones and servers, and to provide high bandwidth to individual users who need it. Instead of sharing a 10Mbps LAN among a number of terminals in a workgroup, a LAN switch can be used, and an individual workstation can get the entire 10Mbps. LAN switches provide great scalability because they enable the network to increase in bandwidth with the fairly simple addition of more switched ports. Thus, LAN switches have many benefits, including scalability in terms of bandwidth, flexibility, and high performance.

Figure 8.9 shows how an Ethernet switch can be used to connect devices that are on the same segment, some of which are served by one shelf of the Ethernet switch and others of which are served by connecting shelves together. On the backplane, you can provide internetworking between the Ethernet segments, so you can provide internetworking on a campuswide basis.

Figure 8.9. An Ethernet switch

As the amount of traffic has grown in the enterprise and as the nature of applications has become more sophisticated, we have been increasing the bandwidth associated with LANs. Today, it is common to see 10Mbps being delivered to an individual desktop and 100Mbps serving as the cluster capacity. To facilitate internetworking between these high-capacity desktops and Fast Ethernet clusters, Gigabit Ethernet is increasingly being used in the backbone. As shown in Figure 8.3 earlier in the chapter, Gigabit Ethernet switches can connect underlying 100Mbps or 10Mbps LAN segments, and the 10Mbps or 100Mbps LAN switches can deliver 10Mbps to the desktop and 100Mbps to the segment.

VLANs

Switched LANs enable us to create VLANs, which don't completely fit the earlier definition of a LAN as being limited in geographical scope. With a VLAN, geography has no meaning. You could have two people in a Singapore office, three in New York, one in London, and four in Cairo, and they could all be part of the same LAN, a VLAN, because the LAN is defined by software rather than by hardware and location. Figure 8.10 shows an example of a VLAN.

Figure 8.10. A VLAN

A switched VLAN is a high-speed, low-latency broadcast group that unites an arbitrary collection of endstations on multiple LAN segments. Switched virtual networking eliminates the bottlenecks that are normally associated with a physical LAN topology by creating high-speed switched connections between endstations on different LAN segments. Users who want to belong to a particular broadcast domain do not have to be physically located on that LAN segment.

VLANs provide a software-based, value-added function by enabling the creation of a virtual broadcast domain, a shared LAN segment within a switched environment. Switching latencies on VLANs are typically one-tenth those of fast routers. However, routers are still required for inter-VLAN communications.

Bridges

Bridges entered the networking scene before routers. Applications for bridges include connecting network segments (for example, by taking 5 to 10 individual clusters and creating the appearance of a single logical VLAN). A bridge can also be used to increase the number of computers on a network or to extend the distance of a segment beyond what the specifications allow. Similarly, a bridge can be used for network segmentation in order to reduce traffic bottlenecks, or to control the flow of network traffic. Bridges can connect similar as well as dissimilar networks.

Bridges have several important functions:

         Learning When the bridge is first connected to the network, it sends an announcement that says, "Hello. I'm your new bridge. What's your address?" All the other devices respond with, "Hello. Welcome to the neighborhood," along with their addresses. The bridge builds a table of local addresses, called the Media Access Control sublayer addresses. The MAC sublayer (which is equivalent to OSI Layer 2) controls access to the shared transmission medium. It is responsible for making the data frames and putting bits in fields that make sense, and it works with the physical layer, Layer 1. MAC standards, including IEEE 802.3, 802.4, and 802.5, define unique frame formats. Every NIC ever made has a globally unique burned-in MAC address.

         Performing packet routing Bridges either filter, ignore, or forward packets.

         Using the Spanning Tree Algorithm Bridges use the Spanning Tree Algorithm to select the most efficient network path and to disable all the other potential routes.

Figure 8.11 illustrates a local bridge installed between two LAN segments that are located at the same local premise. When the bridge is plugged in, it sends out a hello message to its community; the devices answer, and the bridge builds an addressing table. Say that PC A wants to send a document to Printer 1. The bridge realizes that that printer resides within its community. It knows the address and it therefore does not do anything except filter the packet. On the other hand, if PC A is attempting to communicate with Server Z, the bridge says, "Well, I don't know where that server is. It's not part of my local community, so it must be somewhere else on the other side of this bridge." The bridge then broadcasts that information out to as many LAN segments as are connected to the other side of the bridge. In essence, the bridge creates broadcast storms.

Figure 8.11. An example of a local bridge

Bridges are not networkable devices; they can't target a destination network. All they can determine is whether a destination is or is not on its segment, and if the destination is somewhere else, the bridge sends a message to every somewhere else that it knows about. This can be an especially big problem if you use a bridge in a remote mode, as shown in Figure 8.12, because, in essence, you are trying to connect together remote locations by using a WAN link, which is expensive in terms of bandwidth. You pay for every bit sent, so sending messages to LAN segments that don't need to see them across a WAN link that doesn't need to be congested is inefficient.

Figure 8.12. An example of remote bridges

Although bridges can operate in local and remote areas, today they are mostly used in the local environment. They operate at OSI Layer 2 and they are point-to-point they do not understand networking or routing and relaying through a series of nodes. Bridges are protocol independent (Layer 3 and up), which keeps the software simple and inexpensive. Bridges cannot translate between different Layer 2 protocols (for example, between Ethernet and Token Ring). Bridges are primarily used to isolate traffic loads in the local environment because they offer fast throughput; a bridge doesn't have to do intelligent routing, which makes it faster and less expensive than a traditional router. Over time we've merged together the best features of bridges and routers so that some of the problems with each have begun to disappear.

Flat Networks

Flat networks are constructed by using bridges or Layer 2 LAN switches. This type of network is easy to configure, and it promises better performance than hierarchical networks; it offers higher throughput with lower latencies. However, the scalability of a flat network is limited, and a flat network is subject to broadcast storms.

Routers

The most popular internetworking device today is the router (see Figure 8.13). The applications for routers are quite similar to those for bridges. You use them for network segmentation and connection; that is, you use them either to segment larger networks into smaller ones or to connect smaller networks into a larger virtual whole. You can use a router to switch and route packets across multiple communications paths and disparate Layer 2 network types, and because it is a Layer 3 device, a router is networkable it understands how to read network addresses and how to select the destination or target network, so it prohibits broadcast storms. This capability allows routers to act as firewalls between LAN segments. Routers can be associated with traffic filtering and isolation, and because they can read information about the network and transport protocols used, they can make forwarding decisions.

Figure 8.13. An example of routers in a network

Routers can make linking and rerouting decisions, which makes possible selective routing of individual packets over multiple communication paths. Remember that bridges have to disable all but one path, whereas a router can decide on the fly between numerous communications paths. The path a router selects depends on the user's requirements, including cost, speed, and priority.

Routers are protocol specific, but they can support multiple protocols, such as TCP/IP or Novell's IPX/SPX. The key is that each of the protocols a router supports for internetworking requires its own separate routing table, so the more protocols the router supports, the more complex it is, the more memory intensive it is, and the more expensive it is. Routers can be used as bridges to connect similar and dissimilar networks, and again, they are often applied as firewalls.

The functions of a router are as follows:

         Learning A router learns who its neighbors are and builds an addressing table based on their addresses.

         Filtering A router filters packets based on the addressing information it has gathered.

         Routing and switching A router selects the best destinations based on the network addresses, distance, cost, and availability.

         Adapting to network conditions A router adapts to network conditions by changing what it considers to be the best or optimum paths, depending on the network traffic status.

How do routers actually work? A router has input ports for receiving packets and output ports for sending those packets toward their destination. When the packet comes to the input port, the router examines the packet header and checks the destination against a routing table (that is, a database that tells the router how to send packets to various destinations). Based on the information in the routing table, the packet is sent to a particular output port, and the output port sends the packets to a router that is one step closer to that packet's destination. Packets are delivered from node to node (that is, router to router) because, at each node, the router modifies the MAC address to be that of the next node. The destination network address does not change, of course, but the destination MAC changes at each node this is the only way that the packet can travel from one node to the next.

If packets come to the input port more quickly than the router can process them, they are sent to a holding area called an input queue. The router then processes packets from the queue in the order in which were received. If the number of packets received exceeds the length of the queue, packets may be lost. When this happens, an error control mechanism (such as TCP) that is housed on the sending and receiving computers will have the packets resent. (You'll learn more about TCP in Chapter 9, "The Internet: Infrastructure and Service Providers.") There are two types of routing tables:

         Static The simpler kind of routing table is the static routing table. In a static routing table, there are specific ways of routing data to other networks, and only those paths can be used. New routes can be added to the routing table, but they have to be manually programmed. Static routing can't adjust routes as network traffic changes, so it isn't an optimal option for many applications today. A static router knows only its own routing table and does not communicate changes to any of its neighbors.

         Dynamic Dynamic routing is much more useful than static routing. It allows a packet to take multiple routes to reach its final destination. Dynamic routing also allows routers to change the way they route information, based on the amount of network traffic on some paths and routers. In dynamic routing, the routing table is a dynamic routing table that changes as network conditions change. A dynamic router talks with other routers so that they can all update their tables to reflect any changes.

There are two broad types of routing protocols: interior and exterior. Interior routing protocols are typically used only in the routers of an enterprise's intranet (that is, its internal network). Interior routing protocols include Routing Information Protocol (RIP) and Open Shortest Path First (OSPF). Exterior protocols are typically used for routers that are located in the Internet. Whereas there may be many different interior routing schemes, a single exterior routing system manages the whole global Internet, and it's based on Border Gateway Protocol 4 (BGP-4). (Routing protocols are discussed in detail in Chapter 9.)

Hierarchical Networks

Routers, when built into an enterprise internetwork, create a hierarchical network (that is, subnetworks interconnected by routers). They control traffic flow through segmentation, but this can degrade network performance because of delays, and it adds complexity to the overall network configuration. Hierarchical networks are typically used at the edge of a network to interconnect LANs or to provide WAN connectivity to remote LANs. Again, within the customer premises, the simpler bridge-based flat networks are used. As has been mentioned several times, however, traffic levels keep growing, and LANs keep getting busier, so both the edge and the core network are becoming loaded, resulting in network slowness and unacceptable delays.

Routers use a hierarchical addressing scheme, whereby the address includes both the network address and the node address. Routers operate at Layer 3, so they are networkable you can route and relay traffic through a series of routers. Routers are protocol sensitive, so the more internetworking protocols they support, the more complex the software and the greater the number of routing tables and algorithms required to support those protocols.

IP Switches

The network core is responsible for providing interconnectivity, server access, and network management to the edge devices on the network periphery. At the edge of a LAN, a shortage of network capacity, coupled with proliferation of broadcasts and multicasts, can create significant network problems. When the edge demand exceeds the capacity of the core, buffer overruns create capacity overload and lead to lost packets, reducing the availability and reliability of the network. As a result, users today are suffering from congestion, inadequate server access, and slow response times. People want to see information in a matter of a few seconds, so these problems are increasingly frustrating.

The solution to the problem of these increases in traffic in the core and at the edge is the IP switch. The IP switch was designed to speed up increasingly choked networks. An IP switch replaces the slower, more processing-intensive routers. Routers, in general, are slower than switches because they must examine multiple packet fields, make substitutions in the packet headers, and then compute the routes on a packet-by-packet basis. All this activity introduces congestion and latency. The idea behind IP switching was to find a way to make what is essentially a connectionless data technology behave like the more reliable circuit-switched network. The goal is to make networks intranets, extranets, and the Internet faster, as well as to enable the deployment of new genres of applications, including voice, video, and other streaming traffic.

IP switching has two major objectives. One is to add Quality of Service (QoS) support to IP. If we can make a network behave in a connection-oriented fashion, we can allocate resources end-to-end that promise to meet the required service level. (QoS is discussed in detail in Chapter 10, "Next-Generation Networks.") The second objective of IP switching is to provide a way to scale economically because we know that data traffic is growing at a substantial rate (about 30% to 40% per year). IP switching basically replaces a network that consists entirely of Layer 3 hop-by-hop routing and the subsequent associated delays with a route once/switch everything else scenario. That is, the first packet is routed, and then all the subsequent packets are switched on a Layer 2 basis to the destination over the selected virtual circuit. So IP switches vastly improve the performance at LAN/WAN integration points.

Figure 8.14 shows a switched LAN infrastructure that makes use of a high-speed packet backbone, which could be IP, ATM, or MPLS. An Ethernet switch serves a cluster, and a Token Ring switch serves another cluster. An IP switch serves a high-demand cluster. The infrastructure also includes individual servers as well as WAN access devices, all connected to the WAN edge devices, which in this case are the IP switches that help connect or establish a connection-oriented link end-to-end and thereby guarantee latencies and improve the performance associated with the internetwork.

Figure 8.14. A switched LAN infrastructure with an IP/ATM backbone