Table of Contents    Chapter 2    Chapter 4    Netguru.net    Glossary

Chapter 3 - Connectivity

Connectivity refers to the attachment of dissimilar devices together in a network. These devices might include servers, workstations or other key components that allow for the expansion of the network to other networks or resources.

Specifically, this section is designed to give key information on four components that facilitate connectivity - repeaters, bridges, routers and gateways. Each of these devices has its own particular function, and some of these functions can be quite complex. In order to better understand the role and operation of these devices, we would benefit from taking a cursory look at what is known as the OSI model.

The Open Systems Interconnection model was created by the Inter-national Standards Organization in the late 70's. It was to serve as a blueprint for all network communication technologies dividing up all the processes of networking activity into seven distinct layers. The highest layer is the application that a device is running and with which a user may interact. The lowest layer is simply the physical medium of data transfer such as coaxial cable. Each layer has its own distinct functions and services. Here is a summary:

The Open Systems Interconnection Model

We'll spend more time dealing with the OSI Reference Model in a later section of this book. For now, just an overview of its components will do.

The primary purpose for looking at this standard model is provide a framework in which the various connectivity components can be understood, especially within their areas of functionality.

As networks begin to grow and expand, physical limitations are reached. The limitations may have nothing to do with running out of cable or components, but rather running out signal power, or worse yet - running into signal noise. In technical terms this loss of power of a signal is referred to as attenuation while the signal noise is called just that - noise.

In order to minimize these phenomena, special devices called repeaters are incorporated into internetworks (combination of individual networks into larger ones). A repeater does what its name implies. It takes an incoming signal and repeats it, but at a higher power and noise-free.

The repeater is not an amplifier only, as such a device would amplify the good part of the signal as well as the bad. Instead repeaters employ what is known as "signal regeneration". This simply means that the original signal is absorbed, copied and retransmitted along another segment of cabling. This new signal has been beefed up and cleaned up. When it leaves the repeater it is both renewed and noise-free.

Figure 3-1: Repeaters allow us to extend beyond typical distance limitations by regenerating signals.

In reference to the OSI model, the repeater is said to function on the Physical layer. It does not perform any error-checking or repackaging of the original signal that would be viewed as functions of the Data Link layer. Therefore, if a signal had gotten corrupted before reaching the repeater, the device would faithfully pass the corrupted data on to the attached segment.

Signal regeneration takes very little time - generally in the microseconds range. In theory, you could use as many repeaters as you desire, but practically speaking, there are some limits. In Ethernet LANs, a small amount of time is required for a "jamming signal" to make its way completely across the network. If you will recall, the jamming signal is a series of 1 bits designed to alert all workstations attached that a collision has occurred so that no device will transmit while the situation is being remedied. This movement of data across the wiring of a network is called propagation. The propagation time is important to Ethernet, because if the jamming signal does not reach the whole network, two devices on one end may be trying to remedy a collision while the other end of the network doesn't know about and is transmitting as usual. For this reason, most industry experts recommend that a packet of data should not have to pass through more than five repeating devices to reach its destination.

A bridge is a device that is smarter than a repeater. A repeater knows nothing about the data passing through it or the destination of that data. It only knows to regenerate a signal. A bridge on the other hand is informed of where data is going to, and based on that information, can make an intelligent call whether or not to allow the data to go to the destination.

Bridges are able to perform their decision-making because they operate on the Data Link layer of the OSI model. It's on this layer that network systems group packets from data off the wire and make a determination as to where the data goes. Each device on a network has a unique physical station address. This identification is used by devices on network to determine how to send data to one another. A bridge allows two networks to be connected to one another, each having its own group of devices with unique station addresses. The bridge acts as a traffic cop, only allowing data to pass through that is specifically bound from one network to the other. It screens out all data that is transmitted from one device on a network to another device on the same network.

This function is extremely important because it can significantly lower the flow of traffic across a large network. The idea here is to simply divide the network up into smaller networks separated by a bridge thus allowing traffic on one segment to be virtually unaffected by traffic on the other newly created segment. Of course accomplishing this requires a little forethought and planning.

Most companies, when installing bridges, insert them between two geo-graphical segments of network (i.e., between segments servicing two different floors in a building). However, for a bridge to have peak performance, the division would be made based on traffic patterns rather than geographic location of segments. For instance, let's say that a network had begun experiencing traffic problems because of its size and applications. The applications included a word processing package used by the secretarial pool on the first floor and a database package used by accounting and sales staff on the second floor. A contracted network whiz suggests and executes the installation of a bridge between floors. However, much to the whiz's chagrin, the traffic problem persists on the second floor. The reason is simple. Word processing packages like WordPerfect, even when installed on the LAN, do not generate a sizable amount of traffic. When a word processing user initiates the program, a copy of it is placed in their local workstation's memory. Now the only traffic on the LAN created by that user is when or if they edit files stored on a file server, and that only involves periodic transfers of data. The database application, by comparison, generates an enormous amount of data transfers. Queries constantly require data to be accessed on a file server and subsequently transmitted over the LAN. Since the second floor usage was what created the bulk of the load to start with, isolating them with a bridge certainly helped the personnel on the first floor, but left second floor personnel still contending with traffic problems. There are a couple of possible solutions in this case. One would have been to place a bridge between segments so that each had an equal number of database and word processing users. This would have minimized traffic problems for everybody. The other solution is to incorporate another bridge to further segment the LAN on the second floor.

Figure 3-2: Bridges can help control network traffic.

Whatever the case, care should be taken when placing a bridge to provide optimal separation of the two segments. Placing a bridge in the middle of a large LAN with all of the file servers located on one of the segments would be ineffective. As most of the traffic is bound for the file servers, the bridge would allow most of the data to pass through thus allowing a great deal of traffic on the servers' segment. However, placing servers most used by a department on the same segment with them makes more sense, and if the server itself is acting as a bridge (which falls under the capabilities of a NetWare server), then efficiency can be better maintained.

Since networks use different technologies to operate on the Data Link layer, several different types of bridges are found to accommodate such differences. The most common bridge type is the transparent bridge. A transparent bridge builds a table internally as to which segment has which devices. As the devices send data, the bridge adds them to its internal table. Once the bridge has accomplished this, all packets broadcast on a LAN to another node on the same network are discarded. Packets bound for the other LAN are allowed to pass. This simple bridge type requires enough local "intelligence" to be able to create a table and make decisions about it accordingly.

Another type of bridge is called a source routing bridge. This type is incorporated into technology created and used by IBM. In the source routing scenario, each packet contains all the necessary information for routing it to its destination. Therefore the bridge simply forwards the packet to its next destination according to packet information. Obviously a packet bound for another device on the same network would be routed there, thus never crossing the bridge. Source routing makes life simple for a bridge because it does not have to maintain tables for information that the packet already had imbedded into it. The workstation or other device sending the packet has to have all the smarts. In order to embed the correct routing information in a packet, the transmitting device must first know how to get to the intended receiver. This is ascertained through the use of a discovery packet. This special kind of a packet is sent to a destination and may be multiplied as several routes to the destination are encountered. The destination must then reply to every discovery packet. When all the replies reach the source again, the original transmitter determines which route is the best one. It then encodes this data into each packet bound for that destination. The intelligence, once again, lies not at the bridge, but at the source routing device.

Sometimes these two major types of bridges are combined to create what is known as a source routing transparent bridge or (SRT bridge). This bridge looks for the routing information inside each packet associated with source routing. If it doesn't find that type of data, it handles the packet transparently (using internal tables). This type of bridge allows for the easy connection of both source routing and non-source routing networks.

In summary, a bridge functions at the Data Link layer of the OSI model thus allowing it "see" the physical station addresses of each device on the networks attached to it. On the basis of that knowledge, it is able to route data according to information contained in internal tables or data packets themselves. The primary use of a bridge is to isolate network segments so as to reduce traffic flow across the entire network.

Stepping on up the OSI model, we reach the Network layer next. The Network layer allows us to group devices together regardless of whether they share the same physical network or not. We might, for instance, have two distinct LANs in our accounting department, but we might group all of those users as an accounting group by assigning each device in this area a unique logical station address. Then we could refer to the accounting department by way of its logical addresses.

Routers use this type of logical information to perform a very useful task. They are able to determine the best route from a source to a destination regardless of what lies in between. An example would be sending information across the Internet. This huge global network is laden with routers. As we begin sending information over the Internet, each packet is individually directed to the destination. Each time a packet goes through a router, this device attempts to find the best path to send it on closer to its destination. The result is a very dynamic network that can speed data along identifying best paths based on traffic loads and functioning pathways.

Figure 3-3: Routers may serve as boundaries to distinguish networks. Here the router at Network A would choose Path A to send data to Network D because it requires the smallest number of hops (trips through other routers). In fact there are no other routers between Networks A and D .

The methods for determining the best route are many and varied. Modern routers usually incorporate a number of factors in trying determine this type of information. This is necessary because basing a decision on only one factor may prove inefficient. For instance, let's say we are basing our best path decision on selecting the segments along the way with the fastest data throughput. We may end up going through dozens of segments before we reach our destination, thus eliminating our segment speed advantage. Plus, the routers may have selected costly wide area network links, so our packets arrive slowly and our money departs quickly. If we were to choose the best path according to the number of routing devices a packet has to travel through (called hops), we might end up choosing slow or, once again, costly pathways. For these reasons, many routers make a best path decision based on a number of factors, some of which can be weighted subjectively by an administrator.

Routers, due to their sophistication, can be very expensive and relatively slow. The cost of these devices often makes them impractical for small companies. The real benefits come for those companies who have large enterprise-wide networks. If their wide area links are proprietary, they benefit from being able to logically group networks in routing data from one to another in the most efficient way possible. If they are using a public or shared wide area links, then a router provides an extra measure of security, screening out packets that do not belong in a particular logical grouping (including a whole company). This screening capability can deter would-be hackers from getting into a company's networks or prevent electronic junk mail from seeping in off a public link. One small danger when using TCP/IP networks is the possibility of broadcast storms in which a great number of packets inundate a network thus increasing its traffic load and reducing its efficiency. Ironically, the source of broadcast storms is usually misinterpreted router packets. The processing involved in handling routing, regardless of the protocol spoken (like TCP/IP) is extremely CPU intensive. Therefore routing is typically much slower than simple bridging.

Some manufacturers have combined bridges and routers together calling them brouters. These are effectively routers with secondary bridging capabilities built in. A brouter will look for logical station address information in packets that it receives. If that information is unavailable in the packet, the brouters will then simply act as a bridge allowing the packet to pass if its physical station address for the destination qualifies. Often brouters are used to connect different types of LANs together, like token ring and ethernet, while still providing routing services for protocols like TCP/IP. Another deviant from the marriage of router and bridge is a routing bridge used to give a some of the best path selection ability of a router to a bridge instead. These devices are limited as they are not fully functioning routers, only souped up bridges.

We have established that repeaters work on the Physical layer of the OSI model, while bridges function on the Data Link layer and routers on the Network layer. Devices that function at these layers and above to allow interconnection between different network types require a fair amount of sophistication. The changes necessary to create a mainframe-bound message from a PC-based NetWare LAN are significant. The data that is used in the PC world is encoded in a format known as ASCII. IBM host computers use data encoded into a format known as EBCDIC. To switch from one format to another involves the complete restructuring of data. Another thing to consider is that primarily keystroke and screen data are often transmitted along mainframe or minicomputer networks. PC networks can send whole programs and data files, not just terminal data.

Figure 3-4: Gateways enable such diverse systems as PC LANs and mainframe networks to communicate. The gateway typically functions on upper layers of the OSI model.

The sophisticated device required to bridge these two very different environments together is called a gateway. Gateways are unique in that they have the capability of functioning on any level of the OSI model, whatever is necessary to bring together the vastly dissimilar networks. When you purchase a gateway, it is with a certain connection in mind. You might buy one for NetWare and IBM's SNA connections, AppleTalk to DECnet, etc.

Gateways are available in both external and internal models much in the same way that modems are available. External boxes containing the gateway's components tend to be a bit more reliable than their internal plug-in card cousins. Software usually accompanies a gateway, and these devices may be singular in their operation (dedicated) or be multi-functional (non-dedicated).

Connectivity Summary: Internetworking Devices

1. Know the connectivity components and how they operate.

 

2. Know what layer of the OSI model each of the components function on.

 

3. Know how a bridge can be used to reduce traffic problems.

 

4. Know what a router is used for.

 

 

Table of Contents    Chapter 2    Chapter 4    Netguru.net    Glossary

Copyright© 1993 & 1997 by Atrium Technical Inc. All rights reserved. Illegal to copy without written permission of ATI.