Networking Protocols

Now that you have a basic understanding of the OSI model and its related concepts, you can use these OSI concepts to understand how the major protocols work and how each of the protocols within each protocol stack maps to the OSI model, thus describing its function.

In this section you will learn about four major protocol stacks and how each one handles the concepts of addressing, routing, interoperability, and naming. These four protocol stacks are:

• TCP/IP

• IPX/SPX

• NetBEUI

• AppleTalk

TCP/IP

The Transmission Control Protocol/Internet Protocol (TCP/IP) is the protocol of choice today. It, like other protocols, is used to allow two computers to communicate over a network. However, TCP/IP is used not only on local area networks, but also over wide area networks and the Internet. Actually, TCP/IP is the only protocol in use on the Internet. You’ll learn more about TCP/IP in Chapter 3, “TCP/IP Fundamentals.”

Addressing

Addressing the network entities protocol (called hosts in TCP/IP parlance) that runs the TCP/IP is fairly straightforward in TCP/IP. Each host is given (either manually or automatically) a dotted decimal IP address in the format xxx.xxx.xxx.xxx where xxx is a number from 0 to 255. There are several addressing rules, which you will learn more about in Chapter 3.

Because addressing is a Network layer concept, the protocols that deal with addressing can be found at this layer. The Address Resolution Protocol (ARP) is responsible for resolving an IP address to the MAC address of the receiving host. The MAC address is a Data Link layer address hard-coded to each network card at the manufacturer. When a TCP/IP packet is sent, at some point a router will need to determine exactly which station the packet is intended for. On Ethernet networks, the router will use an ARP lookup and ARP broadcasts. Essentially, the router listens for the periodic ARP broadcasts from all hosts and records the information in its ARP cache ( basically a table that says which IP address is associated with which MAC address). When a router receives packets and needs to send them to a particular station on one of its own segments, it examines the IP address of the destination, looks up the MAC address of that station using ARP, and forwards the packet via Ethernet to the intended destination.

Routing

On TCP/IP networks, routing is a fairly involved process. In order to send a packet through an internetwork (like the Internet), the router must have three pieces of information: the IP address of the sender, the IP address of the destination, and the IP address of the next router to which the packet should be sent. The first two are part of the IP datagram being sent, but the router must figure out the last item itself. The router uses information it receives from other routers about what IP addresses they have on their local segments in order to build a logical “map” of the network (called a routing table). Then, the router can determine the best way to get the datagram to its destination and send on the information to the next router.

Routers build routing tables using either RIP or OSPF. The major difference between these two routing protocols (which operate at the Network layer of the OSI model) is that RIP is a distance vector routing protocol and OSPF is a link state routing protocol. The difference between them is pretty simple. Distance vector protocols (like RIP) are used by routers to gather information about the hosts connected to them and to build a table of the addresses and the segment they are on (called a routing table). The router then broadcasts this information to all the routers it is connected to. All of the routers that receive this information add the route information to their own routing tables and rebroadcast them. Approximately every 30 seconds, distance vector protocol routers will rebroadcast their entire routing table. Eventually, all routers in the internetwork know about all the other routers and the networks they serve.

Link state routing protocols (like OSPF) work slightly differently. Whereas RIP routers will broadcast their entire routing table every 60 seconds, a link state router will send out to its “neighbors” only the changes to its routing table. Additionally, link state routers have a more directed relationship with their neighbors. Instead of broadcasting all of its information to everyone on the internetwork, link state protocols (like OSPF) prefer to send only updates and a small amount of information to a specific list of addresses.

Interoperability

Of all the protocols listed in this chapter, no protocol is more flexible or more interoperable than TCP/IP. As the Internet gained popularity, everyone wanted to “get on the Net.” As such, almost every computer had to have two things: a web browser and some form of TCP/IP connection. Therefore, every computer that is connected to the Internet is running TCP/IP in one way or another. Many companies have used the TCP/IP protocol suite to communicate with one another over the Internet.

Additionally, because of this phenomenon, every operating system has some form of TCP/IP protocol stack and, as such, can communicate with other operating systems on some fundamental level.

Naming

TCP/IP hosts are named according to the DNS convention. DNS is a service that resolves names to IP addresses so that we can use friendly names like www.trainsolutions.com to refer to computers instead of unfriendly IP addresses like 192.168.24.31.

There are two parts to a DNS name: the host name (e.g., www) and the domain name (e.g., trainsolutions.com). Each of these components is separated by a period. Typically, you would assign a host name that says what the computer’s function is (i.e., www for a web server). The domain name, on the other hand, is usually the name of the company in which the computer resides, or some related name, followed by .com, .edu, .net, or any other domain suffix. You’ll learn more about DNS in Chapter 3.

IPX/SPX

When Novell NetWare was introduced, it was designed to be a server platform for a local area and wide area networks. To that end, they designed a protocol stack that was very efficient over local area networks and that would also work on wide area networks. That protocol stack was the Internetwork Packet eXchange/Sequenced Packet eXchange, or IPX/SPX.

Addressing

IPX is the Network layer protocol that handles addressing and routing for the IPX/SPX protocol stack. IPX addressing is actually very simple. It takes the 12-digit hexadecimal address, as that is the address for the individual node on that network segment. The network segment is referred to by its own unique 8-digit hexadecimal address. For example, the address:

0001ABF3:12AB341FF414 

would correspond to a station with a MAC address of 12AB341FF414 on the network segment labeled 0001ABF3. Every network segment is assigned its own, unique IPX network address. Since the network card has the MAC address burned in at the factory and, for the most part, can’t be changed, it doesn’t have to be configured. The only configuration that must be done is to assign the IPX network address and configure the server with that address.

Note 

In addition to a station address, routers are given an internal IPX address. This address uniquely identifies a router to the rest of a network. NetWare servers always have an internal IPX address because they can function as routers.

Routing

Most routers that route TCP/IP traffic can also route IPX traffic (although they may require additional software or configuration). IPX/SPX is a routable protocol stack because it has routing protocols designed into it. The routing protocols for IPX/SPX are RIP and NLSP.

IPX RIP is very similar to the RIP protocol in TCP/IP in that RIP for IPX is the distance vector routing protocol for IPX. Similarly, NLSP is the link state routing protocol for IPX/SPX. Both work similarly to their TCP/IP counterparts. RIP uses broadcasts of the entire IPX routing tables to keep all IPX routers updated. And, just like OSPF, NLSP sends out only the changes to the routing tables and then only to a select group of network addresses.

Interoperability

IPX/SPX isn’t as ubiquitous as TCP/IP (which can even be found running on Coke machines), but it holds its own when it comes to allowing many different platforms to talk. Windows 9x, NT, Me, 2000, NetWare, OS/2, and a few versions of Linux come “out of the box” with support for communicating with other entities via the IPX/SPX protocol stack. Before the popularity explosion of the Internet in the mid-1990s, the IPX/SPX protocol stack was the only protocol stack many companies would run.

The only downside to interoperability using IPX/SPX is that many versions of Unix, or other high-end operating systems like OS/400, don’t come with built-in support for the IPX/SPX protocol stack or even with an option for support.

Naming

Really, the only items that have names are the NetWare servers. Generally speaking, you can name a NetWare server anything you want, as long as you follow these rules:

• The name must not include any of the “illegal” characters, including a period (.), a comma (,), a plus sign (+), an equals sign (=), and a backslash (\).

• Names must be less than 64 characters (or 47 characters in older versions of NetWare).

• Names are not case sensitive.

These names are resolved using either Bindery Services or Novell Directory Services. These will be discussed more in Chapter 5, “Major Network Operating Systems.”

NetBEUI

NetBEUI is a Network layer protocol designed to provide support for Net-BIOS networks. NetBIOS is a protocol that was developed by IBM (and later enhanced by Microsoft and Novell) for use with network-aware operating systems like LAN Manager/LAN Server, Windows 9x, Windows NT, and Windows 2000. It is a very fast and efficient protocol with low overhead. Because NetBIOS is small and efficient, it works well on small LANs with between 10 and 200 nodes. The two protocols are often referred to together as NetBEUI/NetBIOS.

Naming and Addressing

There is very little network addressing with NetBEUI/NetBIOS. Actually, for NetBEUI, naming and addressing are the same thing. Each station is configured with a unique name (called the NetBIOS name) that is used for all communications. It’s simple and quick. The only item that must be configured on the workstation is the name of the workstation.

Routing

Because the NetBEUI/NetBIOS protocol stack does not have route discovery protocols and was never designed to be routable, it can’t be routed. All routers will drop NetBEUI/NetBIOS packets. Some routers, however, are smart enough to try and bridge these packets to all segments when it finds out that the packet is NetBEUI.

Interoperability

Only a few operating systems run NetBEUI/NetBIOS. The operating systems for IBM and Microsoft are the primary supporters of this protocol. Windows 9x, NT, 2000, LAN Manager, and OS/2 support NetBEUI/NetBIOS. These operating systems can therefore communicate using NetBEUI/ NetBIOS. The Macintosh operating system, however, does not support NetBEUI natively.

AppleTalk

When Apple introduced the Macintosh in 1984, the Mac included networking software. This networking software used a protocol known as AppleTalk and a cabling system known as LocalTalk. It is a very simple and elegant protocol in that the computer takes care of most of the configuration. You simply plug it in and it works. Because of its simplicity and popularity with Mac users, and because the Mac users wanted a faster version, Apple developed AppleTalk version 2 with support for Ethernet (EtherTalk).

Addressing

Each station on an AppleTalk network uses an address that is 24 bits long. Sixteen of those bits are given to the network, and each network can support 254 nodes. Each network segment can be given either a single 16-bit network number or a range of 16-bit network numbers. If a network is assigned a range of numbers, that network is considered an Extended AppleTalk network because it can support more than 254 nodes. The node address is automatically assigned by the computer itself.

In addition to network numbers, AppleTalk networks use areas called zones. Zones allow an administrator to divide a network into logical areas for easier administration and to make it easier for a user to find resources.

Note 

Although you can have multiple zones on an AppleTalk network, an AppleTalk node can belong to only one zone.

Routing

AppleTalk wasn’t originally designed to be routed over a WAN, but with the release of AppleTalk version 2, Apple included routing functionality with the introduction of the Routing Table Maintenance Protocol (RTMP). RTMP is a distance vector routing protocol, like RIP for both IP and IPX.

Interoperability

The only computer that comes with AppleTalk installed by default is the Macintosh. Most Windows operating systems are able to use the AppleTalk protocol, but require that additional software be installed.

Naming

AppleTalk uses the Name Binding Protocol (NBP) to associate the name of the computer with its network address. It is broadcast based. Every station broadcasts its name when it comes up on a network. The AppleTalk router on a network will cache these names and respond to the NBP request. When a node requests a name resolution, the local router will answer with information it has obtained from this cache.

Note 

If an AppleTalk network doesn’t have a router, each node will perform both NBP requests and NBP responses.