• Tidak ada hasil yang ditemukan

Basic IP Routing

Dalam dokumen TCP/IP Tutorial and Technical Overview (Halaman 113-116)

IP subnet routing

Chapter 3. Routing Protocols

3.1 Basic IP Routing

The fundamental function for routers is present in all IP implementations:

An incoming IP datagram that specifies a destination IP address other than one of the local host's IP address(es), is treated as a normal outgoing IP datagram.

This outgoing IP datagram is subject to the IP routing algorithm (see 2.1.3.4, “IP Routing Algorithm” on page 38) of the local host, which selects the next hop for the datagram (the next host to send it to). This new destination can be located on any of the physical networks to which the intermediate host is attached. If it is a physical network other than the one on which the host originally received the datagram, then the net result is that the intermediate host has forwarded the IP datagram from one physical network to another.

Application TCP

IP Interface Y

Host B

Network Y Network X

Application TCP

IP Interface X

Host A

3376\3376FCK1 IP Routing

Interface X Interface Y Host C Acting as

Router

Figure 69. Router Operation of IP

The normal IP routing table contains information about the locally attached

networks and the IP addresses of other routers located on these networks, plus the networks they attach to. It can be extended with information on IP networks that

are farther away, and can also contain a default route, but it still remains a table with limited information; that is, it represents only a part of the whole IP networks.

That is why this kind of router is called a router with partial routing information.

Some considerations apply to these routers with partial information:

Ÿ They do not know about all IP networks.

Ÿ They allow local sites autonomy in establishing and modifying routes.

Ÿ A routing entry error in one of the routers can introduce inconsistencies, thereby making part of the network unreachable.

Some error reporting should be implemented by routers with partial information via the Internet Control Message Protocol (ICMP) described in 2.2, “Internet Control Message Protocol (ICMP)” on page 58. They should be able to report the following errors back to the source host:

Ÿ Unknown IP destination network by an ICMP Destination Unreachable message.

Ÿ Redirection of traffic to more suitable routers by sending ICMP Redirect messages.

Ÿ Congestion problems (too many incoming datagrams for the available buffer space) by an ICMP Source Quench message.

Ÿ The Time-to-Live field of an IP datagram has reached zero. This is reported with an ICMP Time Exceeded message.

Ÿ Also, the following base ICMP operations and messages should be supported:

– Parameter problem – Address mask – Time stamp

– Information request/reply – Echo request/reply

A more intelligent router is required if:

Ÿ The router has to know routes to all possible IP networks, as was the case for the Internet backbone routers.

Ÿ The router has to have dynamic routing tables, which are kept up-to-date with minimal or no manual intervention.

Ÿ The router has to be able to advertise local changes to other routers.

These more advanced forms of routers use additional protocols to communicate with each other. A number of protocols of this kind exist, and descriptions of the important ones will be given in the following sections. The reasons for this multiplicity of different protocols are basically fourfold:

Ÿ Using Internet terminology, there is a concept of a group of networks, called an autonomous system (AS) (see AS in 3.1.2, “Autonomous Systems” on

page 97), which is administered as a unit. The AS concept arose because the TCP/IP protocols were developed with the ARPANET already in place.

Routing within an AS and routing outside an AS are treated as different issues and are addressed by different protocols.

Ÿ Over two decades several routing protocols were tested in the Internet. Some of them performed well; others had to be abandoned.

Ÿ The emergence of ASs of different sizes called for different routing solutions.

For small to medium-sized ASs a group of routing protocols based upon Distance Vector, such as RIP, became very popular. However, such protocols

do not perform well for large interconnected networks. Link State protocols, such as OSPF, are much better suited for such networks.

Ÿ To exchange routing information between ASs border gateway protocols were developed.

3.1.1 Routing Processes

In TCP/IP software operating systems, routing protocols are often implemented using one of two daemons:3

routed

Pronounced “route D.” This is a basic routing daemon for interior routing supplied with the majority of TCP/IP implementations. It uses the RIP protocol (see 3.3.1, “Routing Information Protocol (RIP)” on page 106).

gated

Pronounced “gate D.” This is a more sophisticated daemon on UNIX-based systems for interior and exterior routing. It can employ a number of additional protocols such as OSPF (see 3.3.4, “Open Shortest Path First (OSPF)” on page 112) and BGP (see 3.4.2, “Border Gateway Protocol (BGP-4)” on page 135).

In TCP/IP hardware implementations, mainly in dedicated router operating systems such as the Common Code for IBM routers or Cisco's Internetworking Operating System (IOS), the routing protocols are implemented in the operating system.

For Multicast Roting Protocols such as DVMRP and MOSPF, please see 9.3,

“Multicast Routing Protocols” on page 472.

3.1.2 Autonomous Systems

The dynamic routing protocols can be divided into two groups:

Ÿ Interior Gateway Protocols (IGPs)

Examples of these protocols are Open Short Path First (OSPF) and Routing Information Protocol (RIP).

Ÿ Exterior Gateway Protocols (EGPs):

An example of these routing protocols is Border Gateway Protocol Verson 4 (BGP-4).

In this book, the term gateway is frequently used to imply an IP router.

Gateway protocols are referred to as interior or exterior depending on whether they are used within or between autonomous systems (ASs).

Interior gateway protocols allow routers to exchange routing information within an AS. Exterior gateway protocols allow the exchange of summary reachability information between separately administered ASs.

3 Daemon, pronounced “demon,” is a UNIX term for a background server process. Usually, daemons have names ending with a d.

An analogous concept for MVS is a server running in a separate address space from TCP/IP; for VM it is a separate service virtual machine, for OS/2 it is a separate OS/2 process, and so on. Although TCP/IP servers are often implemented differently on different platforms, the routed daemon is implemented like this on each of these platforms.

An autonomus system (AS) is defined as a logical portion of larger IP networks that are administered by a single authority. The AS would normally comprise the internetwork within an organization, and would be designated as such to allow communication over public IP networks with ASs belonging to other organizations.

It is mandatory to register an organization's internetwork as an AS in order to use these public IP services (see Figure 70).

Autonomus System A IGPs

Autonomus System B IGPs

Router Router

Autonomus System C IGPs

Router EGP

3376\3376FCK2 Router

Router

Router

Router

Router

Router

Router

Router Router

Figure 70. Autonomous Systems

Figure 70 illustrates three interconnected ASs. It shows that IGPs are used within each AS, and an EGP is used between the three ASs.

Dalam dokumen TCP/IP Tutorial and Technical Overview (Halaman 113-116)