IGRP Concepts

Home à IGRP Items à IGRP Concepts

Desert View IGRP Concepts
Routing Protocols – General Concepts

In Semester 1 chapter 10, we learned that routers are layer 3 devices that are responsible for finding the best path for data to take through the network. Each router maintains a routing table, consisting of destination network addresses and next hop pairs, along with a metric that allows the router to choose which path is the best one to get to the destination network. Therefore, in order to properly route information, the router must have correct information about the topology of the network, as well as a way to determine which path is the best one for forwarding data packets to a given destination. While a network administrator can assign static routes that tell the router which interface to use when sending a packet to a given destination, we learned in Semester 2 chapter 11 that this is is not very practical in most situations, because the routers cannot respond to network changes. Dynamic routing is much more flexible, because it uses routing protocols to allow routers to exchange routing table information periodically, so that they can automatically respond to topology changes and select alternate routes.

A routing protocol tells the router how to send routing table updates, what information is included in those updates, when to send the updates, and to whom to send the updates. They also determine what characteristic of the path are used in calculating the metric of the path. The most common metrics include:

Bandwidth – the data capacity of the link
Delay – the length of time required to move a packet along the link
Load – the amount of activity on the link
Reliability – the error rate of the link
Hop count – the number of routers through which a packet must pass before reaching a destination
Ticks – the delay on the link in IBM PC clock ticks (about 55 milliseconds)
Cost – an arbitrary value assigned to the link by the administrator

There are also three basic classes of routing protocols. The first class is distance vector. Routers running distance vector protocols periodically send full copies of their routing tables to their directly connected neighbors. As information spreads throughout the network, each router learns the direction and rough distance to a destination network, although the router does not know the exact topology of an internetwork. Distance vector routing protocols include RIP and IGRP.

The second class of routing protocol is link-state. Routers running link-state protocols maintain complex databases of network topology information. For this reason, the routers must have greater processor and memory than routers using distance vector protocols. Routers do not send full copies of their routing tables, but instead send only table updates, called link-state advertisements (LSAs), when a link goes down or comes back up. For this reason, link-state protocols use less bandwidth than distance vector, although at initial power-up the network will be flooded with LSAs. Link-state routing protocols include OSPF and NLSP.

The third class of routing protocols is balanced-hybrid, which uses distance-vector metrics, but sends updates only when topology changes, like a link-state protocol. Balanced-hybrid protocols converge rapidly while using less bandwidth and memory than either distance vector or link-state protocols. Protocols include EIGRP and IS-IS.

To be successful, routing protocols must have a mechanism for preventing routing loops. A routing loop occurs when a route goes down, but the information has not yet reached all of the routers across the network.

In the example above, a routing loop could develop if Router C is advertising its path to a route as down, but Router A has not yet received the update and still thinks that a path to the route exists via Router B. Router B could receive the update from C showing the path as down via C, and then receive the update from A and conclude that Router A has a valid alternate path to the route. It could then send that update to C, causing C to conclude that a valid path exists via Router B. The result will be that traffic destined for the down route will be forwarded to Router A, which will then forward it to Router B, which will forward it back to Router A again, and so on.

IGRP
For this network design project, IGRP is specified as the routing protocol. In Semester 3 chapter 5, we learned more specific information about the features and functions of IGRP. IGRP is proprietary to Cisco and was developed to supersede RIP. In contrast to RIP, which only uses hop count as a metric, IGRP uses the metrics of delay and bandwidth by default, and can also use reliability and load. IGRP also gives more importance to bandwidth as a metric, which allows it to select faster links which may have a higher hop counts over slower links with fewer hop counts.

IGRP also scales better than RIP, being designed to provide routing across an entire autonomous system. Some features of IGRP that enhance its ability to avoid routing loops include:

Holddowns – after an update is received about a route, the router will reject any further updates about the route except from the one that originally advertised the update for the duration of the holddown period. This helps to avoid routing loops, although it does increase convergence time.
Split Horizons – a router will not send information about a route back in the same direction that it originally received the update. This provides extra stability on top of holddowns to prevent routing loops.
Poison reverse updates – a router can force a route to be placed into holddown by sending an update putting the metric on the route to infinity. This helps speed convergence across the network while also helping to avoid routing loops.

By default, IGRP sends update broadcasts every 90 seconds. It marks a route as invalid or inaccessible if it does not receive an update about the route within 3 update periods (270 seconds). It flushes the route out of its routing table completely after 7 update periods (630 seconds). IGRP also sends "flash updates" to send updates sooner than the standard update period to notify other routers of a metric change on a route.

IGRP also has a maximum hop count of 255, much higher than the 15 hops allowed by RIP, although by default it is set to 100, and in practice is usually set much lower.

Routing Protocols – Implementation on the Desert View LAN

As mentioned above, IGRP has been specified as the routing protocol for the Washington School District. The topology of the Desert View LAN is relatively simple, with only one route to each of the local networks and one route to the WAN. Since only one possible path exists for traffic to any given destination (as far as the Desert View LAN is concerned) I have chosen to use the default metrics of bandwidth and delay. As soon as the router powers up, it will broadcast its routing table, which should trigger a flash update from the Phoenix N.W. router located in the WAN core. I have also chosen to keep the default values for updates, inaccessible, and flush at 90, 270, and 630 seconds respectively.

To prevent routing loops while keeping convergence time low, holddown timers have been disabled, but split horizons has been enabled. Packets on the District network should not need to pass through more than 4 routers, but hop count has been set to 50 to allow for traffic to the Internet.