CIS 361 Data Communications and Networks

Chapter 11: Network Fundamentals

 

Objectives:

This chapter discusses topics from the third and fourth layers of the ISO-OSI model, having to do with connecting circuits together to make networks. The objectives important to this chapter are:

  • understanding four methods of classifying networks
  • understanding the possible methods of connecting networks
Concepts:
The chapter begins with the statement on page 397 that a network is a collection of circuits connecting nodes. It is helpful to classify networks as being of various types, because each classification method tells us something about the network.

The first classification method is topology, the way the network is shaped and constructed. Six topologies are described:

  • Star - lines are run from nodes to a central point, usually a hub or concentrator, although the central point may be a host computer, as noted on page 400.
  • Hierarchical - a network arranged in layers, where requests have to pass up and down layers to get to various devices
  • Mesh - a system of redundant connections, more fail safe but more costly to wire
  • Bus - a single medium is shared by all nodes, usually through drop cables or taps
  • Ring - traffic must pass from one node to the next in a circular pattern
  • Hybrid - a combination of more than one type
It is important to know the topology used in a network to decide how to service and diagnose problems in it.

Networks can be classified by ownership. A private network is owned by one entity (such as a company) for its own use: outsiders are not meant to intrude. A public network is one that the public is meant to use, like the public telephone network or the Internet. A Value Added Network is one that adds features, for a price, to what is available to the public. For example, any Internet user can access the public features of America Online, but only subscribers to that service can access email through it.

We briefly discussed the three geographic classifications for networks last week:

  1. LAN - Local Area Network, a network in one structure, or in one location smaller than a city
  2. MAN - Metropolitan Area Network, a network that covers the area of a major city, to provide connection services to multiple entities in it.
  3. WAN - Wide Area Network, covers an area larger than a city, as much as the entire world.
Obviously, the geographic terms are relative, and need further discussion. We will talk about LANs in more detail next week.

Networks can also be classified by transmission technology, how they make connections and send messages.

  • Circuit Switching - This is like the PSTN. An actual circuit is created when one node (the caller, on the PSTN) requests a connection to another node. The method described in the text uses a switched circuit. A path is created for that conversation using links that are available. The circuit is in place for these nodes for the duration of their conversation, and ceases to exist as a circuit when the conversation ends. This works well for voice traffic.
  • Packet Switching - Often used with data traffic, this method breaks the messages into packets, small pieces that are numbered and sent across a network to be reassembled at the destination. The thing that makes this packet switching is that the packets can be sent on many different circuits, independent of each other. It is not always done this way, as discussed below. When this is done, it handles more traffic with less waiting over fewer circuits than would be required in the previous scenario. Three submethods are discussed:
    • Switched virtual circuit - this is like circuit switching, in that it lasts for the duration of the conversation
    • Permanent virtual circuit - this is an actual path that is used each time the a circuit is needed between two specific nodes. It is more reliable than the switched version.
    • Datagrams - this is a quick method that does not guarantee the route, arrival order or even the delivery of packets.

Let's continue on page 417 with the discussion of routing messages. Since many paths are possible through many networks, there are several methods of determining the actual path or route to be used. Two major types of routing are discussed on page 417, and it could have been clearer. Try it this way:

  • Connection-oriented - this has more acknowledgments, guaranteed delivery, and may use virtual circuits
  • Connectionless - this will use lots of routes, few acknowledgments, no guarantee, but rapid delivery and rapid retransmission
Three techniques for routing are listed on page 418:
  • Broadcast - all packets are sent to all nodes
  • Centralized - assumes complex paths; a central device keeps tables of what paths to use, for what nodes, so messages are sent only on paths that lead to the destination node
  • Distributed - more devices are used in making the routing decision
There are two ways each of the above techniques could be used: static or dynamic routing. Static routing means one of two things: that only the network administrator can change rules for choosing routes, or that only the routing devices on each end of a circuit can choose routes. Dynamic routing means that devices along the route can make new choices as necessary to keep the data moving.

The next section of the chapter introduces more vocabulary. It is important to know that when you are talking about connecting any two (or more) networks together, you are talking about internetworking. I carefully avoided putting that word at the beginning of a sentence. If I had capitalized it, you could confuse it with Internetworking, which is making connections through the Internet, the network that was originally set up by ARPA, and is now available to the world. Any network that is connected to another network can be referred to as a subnetwork.

In the discussion of TCP/IP on page 420, you should learn that the Internet model is different from the ISO-OSI model, but does the same kind of things, as we have discussed in other chapters. The Internet Protocol (IP) works on the layer 3 of the ISO-OSI model, and does routing and addressing of packets. The Transmission Control Protocol (TCP) works on layer 4 of the ISO-OSI model, and is concerned with error-free, loss-free, connection-oriented delivery of packets. The combination is more reliable than either of the two would be separately.

Mr. Rowe's discussion of the Internet on pages 423-424 is a bit dated, but that is expected in a phenomenon that reinvents itself daily. He mentions four categories of user activity. You should be able to list several others.

It is important for people who do not use the Internet (yet) to become familiar with it, so the vocabulary on the next couple of pages is assigned: telnet, domain name, usenet group, HTML, URL, browser and home page.

Before discussing the next several pages, you should know that the word host has two very different meanings, depending on what kind of network you are talking about. If you are talking about a network of clients and servers, in which the clients request services (like storage space, print service, etc.) the server is called a host. However, if you are talking about an IP network, every device on that has an IP address is called a host. Confusing, isn't it?

Two basic ways of attaching a computer to network are described on pages 426 and 427. The direct connection method resembles the description of a star network from early in this chapter. (Remember that a star can mean a network that uses hubs or concentrators, too.) The other method is to use a Front End Processor, which is a computer that takes on some of the burdens that the host computer would normally have to carry.

The next several pages are not necessary for this course, so we will skip ahead to page 439 and the discussion of two proprietary architectures: SNA and DNA.

Systems Network Architecture (SNA) is an IBM invention. The SNA model is based on a similar but different series of concepts from the ISO-OSI model.

Several layers in it will be familiar and some have new concepts:

  1. Physical - similar to the ISO-OSI layer
  2. Data Link Control - similar to the ISO-OSI layer
  3. Path Control - similar to the ISO-OSI Network layer
  4. Transmission Control- similar to the ISO-OSI Transport layer
  5. Data Flow Control - similar to the ISO-OSI Session layer
  6. Presentation Services - similar to the ISO-OSI layer
  7. Transaction Services - similar to the ISO-OSI Application layer.
SNA divides the network into severaltypes of machines. Think of it as being laid out in 4 Layers:
  • Mainframes
  • Communication Controllers - an interface between the network and the mainframe, also called a Front End Processor (FEP)
  • Cluster Controllers - an interface between the FEP and the LAN segment
  • Terminals - the units on a LAN segment
Other terms are also explained:
  • Host - essentially a server
  • Peripherals - in this terminology, anything attached to the network other than a host or Communications Controller
  • Physical Units - four types are listed on page 439.
  • Logical Units - there are nine types; they differentiate terminal types by their abilities

Digital Network Architecture (DNA) from Digital Equipment Corporation is introduced on page 442. Note the term DECnet, which is used to denote any product that uses DNA.

DECnet supports connecting to other types of networks well. Its model only has five layers in it, but it actually supports all the features of the ISO-OSI model in its five layers.