Routing with IP Technology

This chapter reviews IP routing technology in these sections:

• IP Routing and the OSI Reference Model

• Elements of IP Routing

• IP Routing Transmission Errors

• Routing with Classical IP over ATM

• IP Routing References

An IP router, unlike a bridge, operates at the network layer of the Open Systems Interconnection (OSI) Reference Model. An IP router routes packets by examining the network layer address (IP address). Bridges use data link layer MAC addresses to perform forwarding. See Figure 4-1.

Figure 4-1 OSI Reference Model and IP Routing

When an IP router sends a packet, it does not know the complete path to a destination - only the next hop. Each hop involves three steps:

1 .   The IP routing algorithm computes the next hop IP address and the next router interface, using routing table entries.

2 .   The Address Resolution Protocol (ARP) translates the next hop IP address into a physical MAC address.

3 .   The router sends the packet over the network across the next hop.

IP routers use the following elements to transmit packets:

• IP addresses

• Router interfaces

• Routing tables

• Address Resolution Protocol (ARP)

IP addresses are 32-bit addresses composed of a network part (the address of the network where the host is located) and a host part (the address of the host on that network). See Figure 4-2.

Figure 4-2 IP Address: Network Part and Host Part

IP addresses differ from Ethernet and FDDI MAC addresses, which are unique hardware-configured 48-bit addresses. A central agency assigns the network part of the IP address, and you assign the host part. All devices connected to the same network share the same network part (also called the prefix).

Network Part

The location of the boundary between the network part and the host part depends on the class that the central agency assigns to your network. The three primary classes of IP addresses are A, B, and C:

• Class A address - Uses 8 bits for the network part and 24 bits for the host part. Although only a few Class A networks can be created, each can contain a very large number of hosts.

• Class B address - Uses 16 bits for the network part and 16 bits for the host part.

• Class C address - Uses 24 bits for the network part and 8 bits for the host part. Each Class C network can contain only 254 hosts, but many such networks can be created.

The high-order bits of the network part of the address designate the IP network class.

Subnetwork Part

In some environments, the IP address contains a subnetwork part, at the beginning of the host part of the IP address. Thus, you can divide a single Class A, B, or C network internally, allowing the network to appear as a single network to other networks. The subnetwork part of the IP address is visible only to hosts and gateways on the subnetwork.

When an IP address contains a subnetwork part, a subnet mask identifies the bits that constitute the subnetwork address and the bits that constitute the host address. A subnet mask is a 32-bit number in the IP address format. The 1 bits in the subnet mask indicate the network and subnetwork part of the address. The 0 bits in the subnet mask indicate the host part of the IP address. See Figure 4-3.

Figure 4-3 How a Subnet Mask Is Applied to the IP Address

An example of an IP address that includes network, subnetwork, and host parts is 158.101.230.52 with a subnet mask of 255.255.255.0. This address is divided as follows:

• 158.101 is the network part

• 230 is the subnetwork part

• 52 is the host part

  As shown in this example, the 32 bits of an IP address and subnet mask are usually written using an integer shorthand. This notation translates four consecutive 8-bit groups into four integers ranging from 0 through 255. The subnet mask in the example is written as 255.255.255.0.

A router interface connects the router to a subnetwork. In traditional routing models, the interface is the same as the port because only one interface can exist per port. In the IP routing model for the CoreBuilder 2500 system, more than one port can connect to the same subnetwork.

Each router interface has an IP address and a subnet mask. This router interface address defines both the number of the network to which the router interface is attached and its host number on that network. A router interface IP address serves two functions:

• For sending IP packets to or from the router.

• For defining the network and subnetwork numbers of the segment connected to that interface. See Figure 4-4.

Figure 4-4 Router Interfaces in the CoreBuilder 2500 System

Routing Table

With a routing table, a router or host determines how to send a packet toward its ultimate destination. The routing table contains an entry for every network, subnetwork, and host to which the router or host can forward packets. A router or host uses the routing table when the packet's destination IP address is not on a network or subnetwork to which it is directly connected. The routing table provides the IP address of a router that can forward the packet toward its destination.

The routing table consists of the following elements:

• Destination IP address - The destination network, subnetwork, or host.

• Subnet mask - The subnet mask for the destination IP address.

• Metric - A measure of the distance to the destination. In the Routing Information Protocol (RIP), the metric is the number of hops through routers.

• Gateway - The IP address of the router interface through which the packet travels on its next hop.

• Status - Information that the routing protocol has about the interface.

Figure 4-5 shows the routing table of the router in Figure 4-4.

Figure 4-5 Sample CoreBuilder 2500 Routing Table

Routing table data is updated statically or dynamically:

• Statically - You manually enter static routes in the routing table. Static routes are useful in environments where no routing protocol is used or where you want to override some of the routes that are generated with a routing protocol. Because static routes do not automatically change in response to network topology changes, manually configure only a small number of reasonably stable routes. Static routes do not time out.

• Dynamically - Routers use a protocol such as the Routing Information Protocol (RIP) to automatically exchange routing data. Routes are recalculated at regular intervals. This process helps you to keep up with network changes and allows the system to reconfigure routes quickly and reliably. Interior Gateway Protocols (IGPs), which operate within networks, provide this automated method. The CoreBuilder 2500 system uses RIP and Open Shortest Path First (OSPF) Protocol to configure its routing tables dynamically.

RIP operates using both active and passive devices. Active devices, usually routers, broadcast RIP messages to all devices in a network or subnetwork and update their internal routing tables when they receive a RIP message. Passive devices, usually hosts, listen for RIP messages and update their internal routing tables, but do not send RIP messages.

An active router sends an RIP message every 30 seconds. This message contains the IP address and a metric (distance) from the router to each destination in the router's internal table. In RIP, each router through which a packet must travel to reach a destination counts as one network hop.

OSPF routes packets within and between predefined autonomous systems and areas based on the cost of network links. The OSPF protocol handles network topology changes with a minimum of administrator involvement and routing traffic.

Default Route

In addition to the routes to specific destinations, a routing table can contain a default route. The router uses the default route to forward packets that do not match any other routing table entry. A default route is often used in place of routes to numerous destinations that all have the same gateway IP address and interface number. The default route can be configured statically, or it can be learned dynamically.

Address Resolution Protocol (ARP)

ARP is a low-level protocol used to locate the MAC address corresponding to a given IP address. This protocol allows a host or router to use IP addresses to make routing decisions while it uses MAC addresses to forward packets from one hop to the next.

When the host or router knows the IP address of the next hop toward a packet's destination, the host or router translates that IP address into a MAC address before sending the packet. To perform this translation, the host or router first searches its ARP cache, which is a table of IP addresses with their corresponding MAC addresses. Each device participating in IP routing maintains an ARP cache. See Figure 4-6.

Figure 4-6 Example of an ARP Cache

If the IP address does not have a corresponding MAC address, the host or router broadcasts an ARP request packet to all the devices on the network. The ARP request contains information about the target and source addresses for both the hardware (MAC addresses) and the protocol (IP addresses). See Figure 4-7.

Figure 4-7 Example of an ARP Request Packet

When devices on the network receive this packet, they examine it. If their address is not the target protocol address, they discard the packet. When a device receives the packet and confirms that its IP address matches the target protocol address, the receiving device places its MAC address in the target hardware address field and sends the packet back to the source hardware address. When the originating host or router receives this ARP reply, it places the new MAC address in its ARP cache next to the corresponding IP address. See Figure 4-8.

Figure 4-8 Example of ARP Cache Updated with ARP Reply

After the MAC address is known, the host or router can send the packet directly to the next hop.

Because a router knows only about the next network hop, it is not aware of problems that may be closer to the destination. Destinations may be unreachable if:

• Hardware is temporarily out of service.

• You specified a nonexistent destination address.

• The routers do not have a route to the destination network.

To help routers and hosts discover problems in packet transmission, a mechanism called Internet Control Message Protocol (ICMP) reports errors back to the source when routing problems occur. With ICMP, you can determine whether a delivery failure resulted from a local or a remote problem.

ICMP performs these tasks:

• Tests whether nodes can be reached (ICMP Echo Request and ICMP Echo Reply) - A host or gateway sends an ICMP echo request to a specified destination. If the destination receives the echo request, it sends an ICMP echo reply to the sender. This process tests whether the destination is reachable and responding and verifies that the network's transport hardware and software are working. The ping command is frequently used to invoke this process.

• Creates more efficient routing (ICMP Redirect) - Often the host route configuration specifies the minimum possible routing data that is needed to communicate (for example, the address of a single router). The host relies on routers to update its routing table. In the process of routing packets, a router may detect that a host is not using the best route. The router sends an ICMP redirect to this host, requesting that the host use a different gateway when it sends packets to a destination. The host then sends packets to that destination using the new route.

• Informs sources that a packet has exceeded its allocated time to exist within the network (ICMP Time Exceeded)

CoreBuilder Extended Switching software supports classical IP routing over ATM ARP in an ATM network. The Classical IP over ATM model uses Logical IP Subnetworks (LISs) to forward packets within the network environment.

See the CoreBuilder 2500 Operation Guide for detailed information about the ATM protocol architecture. See the CoreBuilder 2500 Administration Console User Guide for information about how to configure ATM ports.

About Logical IP Subnets (LISs)

A LIS is a group of IP nodes that belong to the same subnetwork and are directly connected to a single ATM network. When you add a node to a LIS through the Administration Console IP interface menu, you define its IP address, subnet mask, and the address of an ATM ARP server that supports it.

ATM ARP Servers

An ATM ARP server maintains a table of IP addresses and their corresponding ATM addresses and circuit information. To forward IP packets over an ATM interface, the network node learns the ATM address for the corresponding IP address from the ATM ARP server.

Each ATM ARP server supports a single LIS. You can associate two or more LISs with the same ATM network, but each LIS operates independently of other LISs on the network.

Several types of network nodes can function as ATM ARP servers:

• Any CoreBuilder system with revision 8.1.0 or later of Extended Switching software

• An ATM switch

• A UNIX workstation

The following sequence describes how the ATM ARP server learns and stores information about the IP and ATM addresses of nodes in the network:

1 .   A node establishes a connection to the ATM ARP server.

2 .   The ATM ARP server sends an inverse ATM ARP request to the node, requesting its IP and ATM address.

3 .   When the node returns this information, the ATM ARP server stores, or caches, it in the ATM ARP server table.

Forwarding to Nodes Within a LIS

Nodes can forward packets directly to other nodes in the same LIS. To forward a packet within the same LIS, the sending node requests a translation from the destination IP address to the corresponding ATM address from the ATM ARP server.

• If the address is known to the server, the server returns a message with this address

• If the address is not known to the server, the server returns a message to advise the sending node that the packet is discarded.

  When the server returns a destination address, the sending node uses this learned address to create a virtual circuit (VC) and to forward this and all subsequent packets to the destination address. The sending node adds this VC to its ATM ARP cache.

Comer, Douglas E. Internetworking with TCP/IP. Volume I: Principles, Protocols, and Architecture. Prentice Hall, Inc., 1991.

Perlman, Radia. Interconnections: Bridges and Routers. Addison-Wesley Publishing Company, Inc., 1992.

Sterns, Richard. TCP/IP Illustrated. Volume 1: The Protocols. Addison-Wesley Professional Computing Services, 1992.

RFC 791. Internet Protocol Specification.

RFC 792. Internet Control Message Protocol Specification.

RFC 1009. Requirements for Internet Gateways.

RFC 1042. A Standard for the Transmission of IP Datagrams over IEEE 802 Networks.

RFC 1058. Routing Information Protocol.

RFC 1122. Requirements for Internet Hosts.

RFC 1577. Classical IP over ATM.

RFC 1583. OSPF Version 2