Networking Technologies

Chapter 8: The OSI Network Layer

 

Objectives:

Chapter 8 is the third chapter about OSI model layers. The objectives important to this chapter are on page 8-1:

1. Understanding the overall purpose of this layer.
2. Identify the uses of the addresses defined at the Network Layer.
3. Identify the characteristics of the three switching methods.
4. Identify the characteristics of the two route discovery methods.
5. Identify the characteristics of the two route selection methods.
6. Identify the characteristics of the connection services that can be implemented at the Network layer.
7. Identify the basic purpose of gateway services.

Concepts:

The overview of the chapter on page 8-2 shows that the Network Layer is concerned with moving data to specific locations across networks. This might have been called the Internetwork Layer, but it was not, so we'll live with it.

The topics and methods for this layer are on page 8-3:

• Network - This layer has 6 topics and is associated with Datagrams
  • Addressing - network addresses. There are two methods:
    • Logical Network
    • Service
  • Switching - route creation for packets, messages and circuits. There are three methods:
    • Packet switching
    • Message switching
    • Circuit switching
  • Route Discovery - finding a route. There are two methods:
    • Distance vector
    • Link-state
  • Route Selection - choosing a route. There are two methods:
    • Static
    • Dynamic
  • Connection Services - flow control, error control and packet sequence control. There are three methods:
    • Network-layer flow control
    • Error control
    • Packet sequence control
  • Gateway Services - There is one method: Network layer translation.

Routers are the hardware most associated with the Network layer.

Addressing is discussed on page 8-4. Network addressing is for sending to and receiving from entities on other networks. Logical Network Addresses are used to distinguish between devices on different networks. This is like the City and State information you put on a letter, more general than the Street Address that goes on the line above it. Every network has a unique ID number. At least it should be unique. If all administrators registered their network numbers, they would be unique. Since registration is not required, unless connecting to the Internet, it is up to the administrators to make sure that each network they connect together has a unique address, allowing the routers to do their jobs.

A Service Address is the logical address that is assigned to each software process, program or protocol that a computer is running. This is necessary because a program may be only one of several programs running on a computer. When that program requests a network service, the service must specifically reply to the process asking for data. So the individual processes are assigned addresses. These addresses can also be called sockets or ports. This kind of address is even more specific than the hardware address of a computer. If the network address is like the city in a postal address, and the hardware address is like the street address, then the service address is like an apartment or suite number within the street address.

The illustration on page 8-6 indicates how a service address is used. Note that the address used to respond to a process includes the Logical Network Address, the Physical Device Address, and the Service Address.

Switching is about the fact that a network may have several possible paths across which to send signals between any two entities. Some path has to be used. The question is how often do we set up a new path? Three methods are discussed:

• Circuit switching - the same path is use for the whole conversation. Long delays are possible to set up the channel, and no one else can use it while two entities have it.
• Message switching - each time a message is sent, it is sent on the best available path. This may not be the same path each time, and messages may arrive out of order. This can be called a store-and-forward network method. Not compatible with real time requirements.
• Packet switching - each packet may travel along a different path, depending on what is available when the packet is sent. There are two approaches to this method:
  • Datagram packet switching - packets often arrive out of order, and must have sequence numbers to reassemble the messages.
  • Virtual circuit packet switching - can be set up for the whole conversation, can vary more often, resembling a point to point connection. Differs from Circuit switching in that packets are still sent individually.

The next topic is Route Discovery. This is the process the router must use to determine what routes exist to reach entities. Two methods are used to discover routes and maintain route tables, so that discovery does not have to start from scratch each time. Route tables keep information about how long routes are, in network terms. Three terms are given to us on page 8-15:

• Hop count - a message makes a hop when it crosses a router. Fewer hops on a route are better.
• Tick count - a tick is one sixteenth of a second. The fewer ticks it takes to get somewhere the better.
• Relative Expense - some paths may cross toll lines, or use channels on which we are charged for use. More commonly, the cost is rated in terms of hops and ticks.

The first method discussed is the Distance Vector Method, known as the bad method. (Think Distance Vector... DV... Darth Vader: bad.) This is a verbose method in which routers communicate with each other, sending their entire tables to each other with each message. This method uses Routing Information Protocol (RIP) to send routing messages. Its advantage is that it is easy to set up and administer. Its disadvantage becomes obvious once you know that routers talk to each other all the time, sending table data to each other, attempting to reach convergence, the state of all routers knowing the information in each others' tables. Convergence takes a lot of traffic and a lot of time using the Distance Vector method.

The second method is the Link-State Method, known as the good method. (Think Link State... LS... Luke Skywalker: good.) This method is less verbose, since the routers only send messages with their whole tables when they first come on line. After that, they send messages about changes in routes to each other, making the messages less frequent and less verbose. Novell uses the NLSP protocol for this method.

Route Selection is done in one of two ways:

• Dynamic Route Selection - the router gathers cost information, and decides on the closest route available. Using this method, each router in a path can adjust the route.
• Static Route Selection - the system administrator (or a device) designates a route to be used. Using this method, routers along the way are not allowed to change the route.

Connection Services are something like the ones in Chapter 7 (Data-Link layer). The same three types exist:

• Unacknowledged Connectionless Services - no flow, error or packet sequence control. No guarantee of delivery.
• Connection-oriented Services - flow, error and packet sequence control. Guaranteed delivery.
• Acknowledged Connectionless Services - uses acknowledgments for flow and error control. No sequence control needed.

The connection services that these processes are used for vary from layer to layer. In the Network layer they are:

• Network-layer flow control - also called congestion control, this is to keep any individual path from being overloaded and overused
• Error control - to request resending of lost packets and discarding of duplicate packets; content errors are handled by the Transport layer. A kind of reliability is attempted.
• Packet sequence control - to put packets in the intended order when received out of order

Finally, Gateway Services are translators. The network layer gateway translates from one set of rules on one network, like Mail Handling Service, to the corresponding set of rules on another network, like Simple Mail Transfer Protocol. The protocols perform similar functions, but a translation is needed when two dissimilar networks are linked together.