Link State Protocols OSPF and Integrated ISIS

Link-state and distance vectors share a common goal—to fill the routing tables with the current best routes. They differ significantly in how they each accomplish the task. The largest difference between the two is that distance vector protocols advertise sparse information; in fact, distance vector protocols know only that other routers exist if the other router broadcasts a routing update to them. When a distance vector protocol in a router hears a routing update, the update says nothing about the routers beyond that neighboring router that sent the update. Conversely, link-state protocols advertise a large amount of topological information about the network, and the routers perform some CPU-intensive computation on the topological data. They even discover their neighbors before bothering to exchange routing information.

To figure out the current best routes, a router processes the link-state topology database using an algorithm called the Dijkstra Shortest Path First (SPF) algorithm. This detailed topology information, along with the Dijkstra algorithm, helps link-state protocols avoid loops and converge quickly.

Link-state protocols prevent loops from occurring easily because each router essentially has a complete map of the network. If you take a trip in your car and you have a map, you are a lot less likely to get lost than someone else who is just reading the signs by the side of the road. Likewise, the detailed topological information helps link-state protocols easily avoid loops. As you will read later, the main reasons that distance vector protocols converge slowly are related to the loop-avoidance features. With link-state protocols, those same loop-avoidance features are not needed, allowing for fast convergence—often in less than 10 seconds.

Open Shortest Path First

OSPF is the most popular link-state IP routing protocol today and is likely to be the most popular one for some time. It works well, is widely deployed, and includes a wide variety of features that have been added over the years to accommodate new requirements.

The basic operation of OSPF differs from that of the distance vector protocols. For the ICND exam, you will need to know a few more details, of course, but for now, a brief look at how OSPF works will help you compare it with distance vector protocols.

One difference relates to how and when OSPF actually sends routing information. A router does not send routing information with OSPF until it discovers other OSPF-speaking routers on a common subnet. The following list gives you some idea of the process:

1. Each router discovers its neighbors on each interface. The list of neighbors is kept in a neighbor table.

2. Each router uses a reliable protocol to exchange topology information with its neighbors.

3. Each router places the learned topology information into its topology database.

4. Each router runs the SPF algorithm against its own topology database to calculate the best routes to each subnet in the database.

5. Each router places the best route to each subnet into the IP routing table.

Link-state protocols do require more work by the routers, but the work is typically worth the effort. A router running a link-state protocol uses more memory and more processing cycles than do distance vector protocols. The topology updates require a large number of bytes to describe the details of every subnet, every router, and which routers are connected to which subnets. However, because OSPF does not send full updates on a regular short interval (like RIP), the overall number of bytes sent for routing information is typically smaller. Also, OSPF converges much more quickly than do distance vector protocols—and fast convergence is one of the most important features of a routing protocol.

OSPF uses a concept called cost for the metric. Each link is considered to have a cost; a route's cost is the sum of the cost for each link. By default, Cisco derives the cost value for a link from the bandwidth, so you can think of the metric as being based on cumulative link bandwidth. (IGRP's metric is based on delay and bandwidth, but it does not treat bandwidth as a cumulative value; it considers only the slowest link in a path.)

The following list points out some of the key features of OSPF:

■ Converges very quickly—from the point of recognizing a failure, it often can converge in less than 10 seconds.

■ Uses short Hello messages on a short regular interval (the Hello interval), with the absence of Hello messages indicating that a neighbor is no longer reachable.

■ Sends partial updates when link status changes, and floods full updates every 30 minutes. The flooding, however, does not happen all at once, so the overhead is minimal.

■ Uses cost for the metric. Integrated IS-IS

Once upon a time, the world of networking consisted of proprietary networking protocols from the various computer vendors. For companies that bought computers from only that one vendor, there was no problem. However, when you used multiple vendor's computers, networking became more problematic.

One solution to the problem was the development of a standardized networking protocol, such as TCP/IP. Skipping a few dozen years of history, you get to today's networking environment, where a computer vendor couldn't sell a computer without it also supporting TCP/IP. Problem solved!

Well, before TCP/IP became the networking protocol standard solving all these problems, the International Organization for Standardization (ISO) worked hard on a set of protocols that together fit into an architecture called Open System Interconnection (OSI). As you recall from Chapter 2, "The TCP/IP and OSI Networking Models," OSI defined its own protocols for Layers 3 through 7, relying on other standards for Layers 1 and 2, much like TCP/IP does today. OSI did not become commercially viable, whereas TCP/IP did—the victory going to the nimbler, more flexible TCP/IP.

So, why bother telling you all this now? Well, OSI defines a network layer protocol called the Connectionless Network Protocol (CLNP). It also defines a routing protocol—a routing protocol used to advertise CLNP routes, called Intermediate System-to-Intermediate System (IS-IS). IS-IS advertises CLNP routes between "intermediate systems," which is what OSI calls routers.

Later in life, IS-IS was updated to include the capability to advertise IP routes as well as CLNP routes. To distinguish it from the older IS-IS, this new updated IS-IS is called Integrated IS-IS. The word integrated identifies the fact that the routing protocol can exchange routing information for multiple Layer 3 routed protocols.

Integrated IS-IS has an advantage over OSPF because it supports both CLNP and IP route advertisement, but most installations could not care less about CLNP. Table 14-5 outlines the key comparison points with all Interior routing protocols for both Integrated IS-IS and OSPF.

Table 14-5 IP Link-State Protocols Compared

Feature

OSPF

Integrated IS-IS

Period for individual reflooding of routing information

30 minutes

15 minutes

Metric

Cost

Metric

Supports VLSM

Yes

Yes

Convergence

Fast

Fast

0 0

Post a comment