hometocprevnextglossaryfeedbacksearchhelp

Table of Contents

The Mechanics of Routing Protocols

The Mechanics of Routing Protocols

The previous chapters examined what routers do, how they can be used, and their various physical mechanisms. This chapter gives you a closer look at how they operate and describes the two primary types of routing: static and dynamic. Of these two, only dynamic uses routing protocols. As a result, dynamic routing is much more powerful and complicated.

Dynamic routing protocols are the technology that enables routers to perform some of their more vital functions. This includes discovering and maintaining routes, converging on an agreement of a network's topology, as well as some of the differences in the ways to calculate routes. Examining the mechanics of these basic functions will provide the context for a more in-depth examination of each of the more commonly encountered dynamic routing protocols, which is provided in Part III, "Routing Protocols."

Routing

Routers can route in a two basic ways. They can use preprogrammed static routes, or they can dynamically calculate routes using any one of a number of dynamic routing protocols. Dynamic routing protocols are used by routers to perform discover routes. Routers then mechanically forward packets (or datagrams) over those routes.

Statically programmed routers cannot discover routes; they lack any mechanism to communicate routing information with other routers. Statically programmed routers can only forward packets using routes defined by a network administrator.

In addition to static programming of routes, there are three broad categories of dynamic routing protocols:

The primary differences between these types of dynamic routing protocols lie in the way that they discover and calculate new routes to destinations.


Note Two Functional Classes of Dynamic Routing Protocols

Routing protocols can
be classified in many ways, including by many of their operational characteristics such as their field of use, the number of redundant routes to each supported destination, and so on. This book classifies them by the way that they discover and calculate routes. However, it is still useful to reference routing protocols by their field of use. In other words, categorize them by the role that they perform in an internetwork. There are two functional classes of dynamic routing protocols, for example: Interior Gateway Protocols (IGPs) and Exterior Gateway Protocols (EGPs).

Perhaps the easiest way to explain this is that IGPs are used within autonomous systems, such as intranets, whereas EGPs are used between autonomous systems. Consequently, the Border Gateway Protocol (BGP---an EGP) is the protocol used to calculate routes across the Internet. The Internet, from a routing perspective, is nothing more than a backbone transport for a global collection of privately owned and operated autonomous systems.

The specific protocols covered in Part III are identified as either IGPs or EGPs.

Static Routing

The simplest form of routing is preprogrammed and, consequently, static routes. The tasks of discovering routes and propagating them throughout a network are left to the internetwork's administrator(s).

A router programmed for static routing forwards packets out of predetermined ports. After the relationship between a destination address and a router port is configured, there is no longer any need for routers to attempt route discovery or even communicate information about routes.

There are many benefits to using static routes. For instance, statically programmed routes can make for a more secure network. There can be only a single path into, and out of, a network connected with a statically defined route. That is, of course, unless multiple static routes are defined.

Another benefit is that static routing is much more resource efficient. Static routing uses far less bandwidth across the transmission facilities, doesn't waste any router CPU cycles trying to calculate routes, and requires far less memory. In some networks, you might even be able to use smaller, less expensive routers by using static routes. Despite these benefits, you must be aware of some inherent limitations to static routing.

Drawbacks to Static Routing

In the event of a network failure, or other source of topology change, the onus is on the network administrator to manually accommodate the change. Figure 7-1 illustrates this point.


Figure 7-1: A simple internetwork with static routes.


In this simple example, the networks' administrators have collaborated on a route redistribution scheme that they believe will minimize their workload as well as network traffic loads. The internetwork is relatively small, consisting of three different networks, one of which supports a stub network. Each network uses its own address space and a different dynamic routing protocol. Given the innate incompatibility of the three different routing protocols, the administrators chose not to redistribute routing information among their networks. Rather, they aggregated the routes into network numbers, and statically defined paths for them. Table 7-1 summarizes the routing tables of the three gateway routers. Router D connects a small, stub network to the other networks. As such, this router uses its serial port as a default gateway for all packets addressed to any IP address that does not belong to 192.168.126.


Note For more information on redistributing routing information among dissimilar routing protocols, refer to "Internetworking with Dissimilar Protocols."

Table 7-1: Statically Defined Routes

Router Destination Next Hop

A

172.16.0.0

B

A

192.168.0.0

C

B

10.0.0.0

A

B

192.168.0.0

C

C

10.0.0.0

A

C

172.16.0.0

B

C

192.168.126.0

D

In this scenario, Router A forwards all packets addressed to any hosts within the 172.16 network address space to Router B. Router A also forwards all packets addressed to hosts within network 192.168 to Router C. Router B forwards all packets addressed to any hosts within the 192.168 address space to Router C. Router B forwards packets addressed to hosts within Network 10 to Router A. Router C forwards all packets destined for Network 10 to Router A, those packets destined for 172.16 to Router B. Additionally, Router C forwards packets addressed to 192.168.126 to Router D, its stub network. This network is a stub because it is literally a dead end in the network. There is only one way in---and one way out. This small network depends completely on its link to Router C, and Router C itself, for connectivity to all the internetworked hosts.

In this example, a failure will result in unreachable destinations despite the fact that an alternative path is available for use. In Figure 7-2, the transmission facility between Gateway Routers A and C has failed.


Figure 7-2: A link failure in a statically programmed internetwork can disrupt communications.


The effect of this failure is that end systems in networks 10 and 192.168 cannot communicate with each other, even though a valid route exists through Router B! Table 7-2 summarizes the effects of this type of failure on the routing tables.


Table 7-2:
Static Routes with a Failed Link

Router Destination Next Hop

A

172.16.0.0

B

A

192.168.0.0

C - Unreachable

B

10.0.0.0

A

B

192.168.0.0

C

C

10.0.0.0

A - Unreachable

C

172.16.0.0

B

The lack of any dynamic mechanism prevents Routers A and C from recognizing the link failure. They are not using a routing protocol that would otherwise discover and test the qualities of the links to known destinations. Consequently, they cannot discover the alternative path through Router B. Although this is a valid and usable route, their programming prevents them from discovering it or using it. This situation will remain constant until the network administrator takes corrective action manually.

What's Static Routing Good For?

At this point, you might be wondering what possible benefit there might be in statically defined routes. Static routing is good only for very small networks that only have a single path to any given destination. In such cases, static routing can be the most efficient routing mechanism because it doesn't consume bandwidth trying to discover routes or communicate with other routers.

As networks grow larger, and add redundant paths to destinations, static routing becomes a labor-intensive liability. Any changes in the availability of routers or transmission facilities in the WAN must be manually discovered and programmed in. WANs that feature more complex topologies that offer multiple potential paths absolutely require dynamic routing. Attempts to use static routing in complex, multipath WANs will defeat the purpose of having that route redundancy.

At times, statically defined routes are desirable, even in large or complex networks. Static routes can be configured to enhance security. Your company's connection to the Internet could have a statically defined route to a security server. No ingress would be possible without having first passed whatever authentication mechanisms the security server provides.

Alternatively, statically defined routes might be extremely useful in building extranet connections using IP to other companies that your employer does a lot of business with. Finally, static routes might be the best way to connect small locations with stub networks to your WAN. The point is that static routes can be quite useful. You just need to understand what they can and can't do.

Distance-Vector Routing

In routing based on distance-vector algorithms, also sometimes called Bellman-Ford algorithms, the algorithms periodically pass copies of their routing tables to their immediate network neighbors. Each recipient adds a distance vector---that is, its own distance "value"---to the table and forwards it on to its immediate neighbors. This process occurs in an omnidirectional manner among immediately neighboring routers. This step-by-step process results in each router learning about other routers and developing a cumulative perspective of network "distances."


Note Network distances and costs are somewhat euphemistic in this context. They do not relate directly to either physical distances or monetary costs.

The cumulative table is then used to update each router's routing tables. When completed, each router has learned vague information about the "distances" to networked resources. It does not learn anything specific about other routers, or the network's actual topology.

Drawbacks to Distance-Vector Routing

Under certain circumstances, distance-vector routing can actually create routing problems for distance-vector protocols. A failure or other change in the network, for example, requires some time for the routers to converge on a new understanding of the network's topology. During the convergence process, the network may be vulnerable to inconsistent routing, and even infinite loops. Safeguards can contain many of these risks, but the fact remains that the network's performance is at risk during the convergence process. Therefore, older distance-vector protocols that are slow to converge may not be appropriate for large, complex WANs.

Even in smaller networks, distance-vector routing protocols may be problematic at worst, or suboptimal at best. This is because the simplicity that is this genre's strength can also be a source of weakness. Figure 7-3 presents an internetwork with specific geographic locations.

In this example, Network 1 is in New York, Network 2 is in Seattle, Network 3 is in Philadelphia, and Network 4 is in Minneapolis. The distance-vector routing protocol uses a statically assigned cost of 1 for each hop, regardless of the distance of the link or even its bandwidth. Table 7-3 summarizes the number of hops to each of the destination network numbers. Notice that the routers do not have to create separate entries in their routing tables for every known end system. They can do so, but instead usually summarize routes. Route summarization is the truncation of the host portion of an IP address; only the network address is stored. In theory, the same path can be used to get to all the hosts or end systems on any given network. Therefore, nothing is gained by creating separate entries for each host address.


Figure 7-3:
An internetwork using a distance-vector routing protocol.



Table 7-3:
The Number of Hops with the Distance-Vector Protocol

Router Destination Next Hop Number of Hops to Destination

A

172.16.0.0

B

1

A

192.168.125.0

C

1

A

192.168.253.0

B or C

2

B

10.0.0.0

A

1

B

192.168.125.0

C

1

B

192.168.253.0

D

1

C

10.0.0.0

A

1

C

172.16.0.0

B

1

C

192.168.253.0

D

1

D

10.0.0.0

B or C

2

D

172.16.0.0

B

1

D

192.168.125.0

C

1

In any internetwork with redundant routes, it is better to use a distance-vector protocol than to use static routes. This is because distance-vector routing protocols can automatically detect and correct failures in the network. Unfortunately, they aren't perfect. Consider the routing table entries for Gateway Router A, for example. This is the New York gateway. From its perspective, the Minneapolis gateway is two hops away, regardless of whether it goes through Philadelphia or Seattle. In other words, this router would be indifferent to accessing Minneapolis through either Philadelphia or Seattle.

If all the variables in the network were held constant (including such things as traffic levels, the bandwidth of each link, and even transmission technology), the geographically shortest path would incur the least amount of propagation delay. Therefore, logic dictates taking the shorter route, through Philadelphia. In reality, such logic is beyond the capabilities of simple distance-vector protocols. Distance-vector protocols aren't exactly limited by this because propagation delay is often the least significant of the factors driving the performance of a route. Bandwidth and traffic levels can both have much more noticeable effects on the performance of a network.

What's Distance-Vector Routing Good For?

Generally speaking, distance-vector protocols are very simple protocols that are easy to configure, maintain, and use. Consequently, they prove quite useful in very small networks that have few, if any, redundant paths and no stringent network performance requirements. The epitome of the distance-vector routing protocol is Routing Information Protocol (RIP). RIP uses a single distance metric (cost) to determine the best next path to take for any given packet. RIP has been widely used for decades, and has only recently warranted updating. For more information on RIP, refer to "Routing Information Protocol," and "Routing Information Protocol Version 2."

Link-State Routing

Link-state routing algorithms, known cumulatively as shortest path first (SPF) protocols, maintain a complex database of the network's topology. Unlike distance-vector protocols, link-state protocols develop and maintain a full knowledge of the network's routers as well as how they interconnect. This is achieved via the exchange of link-state advertisements (LSAs) with other routers in a network.

Each router that has exchanged LSAs constructs a topological database using all received LSAs. An SPF algorithm is then used to compute reachability to networked destinations. This information is used to update the routing table. This process can discover changes in the network topology caused by component failure or network growth.

In fact, the LSA exchange is triggered by an event in the network, instead of running periodically. This can greatly expedite the convergence process because there is no need to wait for a series of arbitrary timers to expire before the networked routers can begin to converge!

If the internetwork depicted in Figure 7-3 were to use a link-state routing protocol, the concerns about connectivity between New York and Minneapolis would be rendered moot. Depending on the actual protocol employed, and the metrics selected, it is highly likely that the routing protocol could discriminate between the two paths and try to use the best one. Table 7-4 summarizes the contents of the gateways' routing tables.


Table 7-4:
Hop Counts in a Link-State Network

Router Destination Next Hop Number of Hops to Destination

A

172.16.0.0

B

1

A

192.168.125.0

C

1

A

192.168.253.0

B

2

A

192.168.253.0

C

2

B

10.0.0.0

A

1

B

192.168.125.0

C

1

B

192.168.253.0

D

1

C

10.0.0.0

A

1

C

172.16.0.0

B

1

C

192.168.253.0

D

1

D

10.0.0.0

B

2

D

10.0.0.0

C

2

D

172.16.0.0

B

1

D

192.168.125.0

C

1

As is evident in this table's routing entries for the New York-to-Minneapolis routes, a link-state protocol would remember both routes. Some link-state protocols may even provide a means to assess the performance capabilities of these two routes, and bias toward the better-performing one. If the better- performing path, for example the route through Philadelphia, were to experience operational difficulties of any kind (including congestion or component failure), the link-state routing protocol would detect this change and begin forwarding packets through Seattle.

Drawbacks to Link-State Routing

Despite all its features and flexibility, link-state routing raises two potential concerns:

These are hardly fatal flaws in the link-state approach to routing. The potential performance impacts of both can be addressed, and resolved, through foresight, planning, and engineering.

What's Link-State Routing Good For?

The link-state approach to dynamic routing can be quite useful in networks of any size. In a well-designed network, a link-state routing protocol will enable your network to gracefully weather the effects of unexpected topological change. Using events, such as changes, to drive updates (rather than fixed-interval timers) enables convergence to begin that much more quickly after a topological change.

The overheads of the frequent, time-driven updates of a distance- vector routing protocol are also avoided. This allows more bandwidth to be used for routing traffic rather than for network maintenance, provided you design your network properly.

A side benefit of the bandwidth efficiency of link-state routing protocols is that they facilitate network scalability better than either static routes or distance-vector protocols. When juxtaposed with their limitations, it is easy to see that link-state routing is best in larger, more complicated networks or in networks that must be highly scalable. It may be challenging to initially configure a link-state protocol in a large network, but is well worth the effort in the long run. For more information on link-state routing, refer to "Open Shortest Path First."

Hybridized Routing

The last form of routing discipline is hybridization. The balanced hybrid routing protocols use distance-vector metrics but emphasize more accurate metrics than conventional distance-vector protocols. They also converge more rapidly than distance-vector protocols but avoid the overheads of link-state updates. Balanced hybrids are event driven rather than periodic and thereby conserve bandwidth for real applications.

Although "open" balanced hybrid protocols exist, this form is almost exclusively associated with the proprietary creation of a single company, Cisco Systems, Inc. Its protocol, Enhanced Interior Gateway Routing Protocol (EIGRP), was designed to combine the best aspects of distance-vector and link-state routing protocols without incurring any of their performance limitations or penalties. Given that this class of dynamic routing protocol is dominated by EIGRP, its benefits and limitations are examined in more detail in "Enhanced Interior Gateway Routing Protocol," rather than examined generically in this chapter.

Performance Characteristics of Hybridized Routing

One of the most difficult, yet critical, tasks that must be surmounted when building an internetwork is the selection of a routing protocol. As the preceding sections indicate, a rich palette of options await the prospective internetwork architect. Although the preceding overviews should help you to differentiate among the various classes of dynamic routing protocols, this is just the beginning. You still need to select a specific routing protocol, or protocols, from the variety that may be available in each class.

One of the best ways to start narrowing down the list of potential protocols is by evaluating each protocol's performance characteristics relative to projected requirements. Unlike hardware, you can't just compare routing protocols' packets-per-second or bandwidth ratings. They don't exist! Instead, you should look at how effectively each protocol performs the various tasks that support internetworking.

Two of the most important of these tasks are convergence and route calculation. The remaining sections of this chapter examine each of these concepts in more detail and should adequately prepare you for the more detailed examinations of individual protocols in the following chapters.


Note Of course, many other performance attributes must be considered, including maximum network diameter and how well a given protocol accommodates heavy traffic loads. Such characteristics, however, tend to be more applicable to specific protocols than to the three classes of dynamic routing protocols identified earlier in this chapter. As such, these performance characteristics are examined individually in Part III.

Convergence

One of the most fascinating aspects of routing is a concept known as convergence. Quite simply, whenever a change occurs in a network's topology, or shape, all the routers in that network must develop a new understanding of what the network's topology is. This process is both collaborative and independent; the routers share information with each other, but must independently calculate the impacts of the topology change on their own routes. Because they must mutually develop an agreement of the new topology independently from different perspectives, they are said to converge on this consensus.

Convergence is necessary because routers are intelligent devices that can make their own routing decisions. This is simultaneously a source of strength and vulnerability. Under normal operating conditions, this independent and distributed intelligence is a source of tremendous advantage. During changes in the network's topology, the process of converging on a new consensus of the network's shape may actually introduce instability and routing problems.

Accommodating Topological Changes

Unfortunately, the independent nature of routers can also be a source of vulnerability whenever a change occurs in the network's topology. Such changes, by their very nature, change a network's topology. Figure 7-4 illustrates how a change in the network is, in fact, a change in its topology.


Figure 7-4: A four-gateway internetwork.


Figure 7-4 features another fairly simple, four-node internetwork with some route redundancy. Table 7-5 summarizes the routing tables of the four routers. For the sake of this example, consider this table to be preconvergence routing table information.


Table 7-5:
Router Destination Next Hop Number of Hops to Destination

A

172.16.0.0

B

1

A

192.168.125.0

C

1

A

192.168.253.0

B or C

2

B

10.0.0.0

A

1

B

192.168.125.0

A or D

2

B

192.168.253.0

D

1

C

10.0.0.0

A

1

C

172.16.0.0

A or D

2

C

192.168.253.0

D

1

D

10.0.0.0

B or C

2

D

172.16.0.0

B

1

D

192.168.125.0

C

1

Preconvergence Routing Table Contents

If packets sent by Router C to Server 192.168.253.2 suddenly become undeliverable, it is likely that an error occurred somewhere in the network. This could have been caused by a seemingly infinite number of different, specific failures. Some of the more common suspects include the following:

Obviously, the new network topology can't be determined until the exact location of the failure has been identified. Similarly, the routers cannot attempt to route around the problem until the failure location has been isolated. If either of the first two scenarios occurred, server 192.168.253.2 would be completely unavailable to all the users of the internetwork, regardless of any route redundancy that may have been built into the network.

Similarly, if router D had failed completely, all the LAN-attached resources at that location would be isolated from the rest of the network. If the failure was either a partial failure of that router, or elsewhere in the network, however, there might still be a way to reach Server 192.168.253.2. Finding a new route to 192.168.253.2 requires the network's routers to recognize and agree on which piece of the network failed. In effect, subtracting this component from the network changes the network's topology.

To continue with the example, assume that Router D's serial interface port to router C has failed. This renders the link between C and D unusable. Figure 7-5 illustrates the new network topology.


Figure 7-5: The link between Routers C and D is unusable.


Routers using a dynamic routing protocol would quickly determine that Server 192.168.253.2 was unreachable through their current, preferred route. Individually, none of the routers could determine where the actual failure occurred, nor could they determine whether any viable alternative routes still existed. By sharing information with each other, however, a new composite picture of the network can be developed.


Note For the purposes of this chapter, this example uses an intentionally generic method of convergence. More specific details about each routing protocol's convergence characteristics are presented in Part III.

The routing protocol used in this internetwork is relatively simple. It limits each router to exchanging routing information with its immediate neighbors, although it supports the recording of multiple routes per destination. Table 7-6 summarizes the pairs of immediately adjacent routers illustrated in Figure 7-5.


Table 7-6: Routers that Share Routing Information with Immediate Neighbors

Router A B C D

A

---

Yes

Yes

No

B

Yes

---

No

Yes

C

Yes

No

---

Yes

D

No

Yes

Yes

---

The entries in Table 7-6 that contain the word Yes indicate a physically adjacent pair of routers that would exchange routing information. The entries that contain a dash denote the same router: A router cannot be adjacent to itself. Finally, those entries that contain the word No indicate nonadjacent routers that cannot directly exchange routing information. Such routers must rely on their adjacent neighbors for updates about destinations on nonadjacent routers.

From this table, it is apparent that because they are not directly connected to each other, Routers A and D must rely on Routers B and C for information about each other's destinations. Similarly, Routers B and C must rely on Routers A and D for information about each other's destinations.

Figure 7-6 shows this sharing of routing information between immediate neighbors.


Figure 7-6: Immediate neighbors sharing routing data.


The important implication in this scenario is that, because not every router is immediately adjacent to every other router, more than one routing update may be required to fully propagate new routing information that accommodates the failed link. Therefore, accommodating topological change is an iterative and communal process.

For the sake of simplicity, assume that convergence occurs within two routing table updates in this example. During the first iteration, the routers are starting to converge on a new understanding of their topology. Routers C and D, because of the unusable link between them, cannot exchange routing information. Consequently, they invalidate this route and all destinations that use it. Table 7-7 summarizes the contents of the four routers' routing tables during the convergence process. Note that the contents of some routing tables may reflect the mistaken belief that the link between Routers C and D is still valid


Table 7-7: Midconvergence Routing Table Contents

.
Gateway Router Destination Next Hop Number of Hops to Destination

A

172.16.0.0

B

1

A

192.168.125.0

C

1

A

192.168.253.0

B or C

2

B

10.0.0.0

A

1

B

192.168.125.0

A or D

2

B

192.168.253.0

D

1

C

10.0.0.0

A

1

C

172.16.0.0

A only (D failed)

2

C

192.168.253.0

D - Invalid route

Not reachable

D

10.0.0.0

B or C

2

D

172.16.0.0

B

1

D

192.168.125.0

C - Invalid route

Not reachable

In Table 7-7, Routers C and D have invalidated the route between them. Routers A and B, however, still believe that their routes through this link are viable. They must await a routing update from either Router C and/or D before they can recognize the change in the internetwork's topology.

Table 7-8 contains the contents of the four routers' routing tables after they have converged on a new topology. Remember that this is an intentionally generic depiction of the convergence process; it is not indicative of any particular routing protocol's mechanics.


Table 7-8: Postconvergence Routing Table Contents

Router Destination Name Next Hop Number of Hops to Destination

A

172.16.0.0

B

1

A

192.168.125.0

C

1

A

192.168.253.0

B only

2

B

10.0.0.0

A

1

B

192.168.125.0

A only

2

B

192.168.253.0

D

1

C

10.0.0.0

A

1

C

172.16.0.0

A only

2

C

192.168.253.0

A

3

D

10.0.0.0

B only

2

D

172.16.0.0

B

1

D

192.168.125.0

B only

3

As evident in Table 7-8, all the routers in the internetwork eventually agree that the link between C and D is unusable, but that destinations in each autonomous system are still reachable via an alternative route.

Convergence Time

It is virtually impossible for all routers in a network to simultaneously detect a topology change. In fact, depending on the routing protocol in use, as well as numerous other factors, a considerable time delay may pass before all the routers in that network reach a consensus, or agreement, on what the new topology is. This delay is referred to as convergence time. The important thing to remember is that convergence is not immediate. The only uncertainty is how much time is required for convergence to occur.

Some factors that can exacerbate the time delay inherent in convergence include the following:

The effects of some of these factors can be minimized through careful network engineering. A network can be engineered to minimize the load on any given router or communications link, for example. Other factors, such as the number of routers in the network, must be accepted as risks inherent in a network's design. It may be possible, however, to engineer the network such that fewer routers need to converge! By using static routes to interconnect stubs to the network, you reduce the number of routers that must converge. This directly reduces convergence times. Given these factors, it is clear that the two keys to minimizing convergence times are

Route Calculation

As demonstrated through the examples in the preceding section, convergence is absolutely critical to a network's capability to respond to operational fluctuations. The key factor in convergence is communications among the routers in the network. Routing protocols are responsible for providing this function. Specifically, these protocols are designed to enable routers to share information about routes to the various destinations within the network.


Note Route Flapping

One symptom of network instability that may arise is known as route flapping. Route flapping is just the rapid vacillation between two, or more, routes. Flapping happens during a topology change. All the routers in the network must converge on a consensus of the new topology. Toward this end, they begin sharing routing information.

In an unstable network, a router (or routers) may be unable to decide on a route to a destination. Remember that during convergence a router may alter its primary route to any given destination as a result of the last- received update. In complex, but unstable networks with redundant routes, a router may find itself deciding on a different route to a given destination every time it receives an update. Each update nullifies the previous decision and triggers another update to the other routers. These other routers, in turn, adjust their own routing tables and generate "new" updates. This vicious cycle is known as flapping. You may find it necessary to power down affected routers and slowly develop convergence in your network, one router at a time.

Unfortunately, all routing protocols are not created equal. In fact, one of the best ways to assess the suitability of a routing protocol is to evaluate its capabilities to calculate routes and converge relative to other routing protocols. It should be obvious from the previous list of factors that convergence times may be difficult for you to calculate with any degree of certainty. Your router vendor may be able to assist you with this process, even if the vendor provides you with general estimates only.

A routing protocol's convergence capability is a function of its capability to calculate routes. The efficacy of a routing protocol's route calculation is based on several factors:

Storing Multiple Routes

Some routing protocols attempt to improve their operational efficiency by only recording a single route (ideally, the best route) to each known destination. The drawback to this approach is that when a topology change occurs, each router must calculate a new route through the network for the impacted destinations.

Other protocols accept the processing overheads that accompany larger routing table sizes and store multiple routes to each destination. Under normal operating conditions, multiple routes enable the router to balance traffic loads across multiple links. If, or when, a topology change occurs, the routers already have alternative routes to the impacted destinations in their routing tables. Having an alternative route already mapped out does not necessarily accelerate the convergence process. It does, however, enable networks to more gracefully sustain topology changes.

Initiating Updates

As you will see in Part III, some protocols use the passage of time to initiate routing updates. Others are event driven. That is, they are initiated whenever a topological change is detected. Holding all other variables constant, event-driven updates result in shorter convergence times than timed updates.

Timed Updates

A timed update is a very simple mechanism. Time is decremented in a counter as it elapses. When a specified period of time has elapsed, an update is performed regardless of whether a topological change has occurred. This has two implications:

Event-Driven Updates

Event-driven updates are a much more sophisticated means of initiating routing updates. Ostensibly, an update is initiated only when a change in the network's topology has been detected. Given that a topology change is what creates the need for convergence, this approach is obviously the more efficient one.

You can select an update initiator just by selecting a routing protocol; each protocol implements either one or the other. Therefore, this is one factor that must be considered when selecting a routing protocol.

Routing Metrics

The routing protocol determines another important mechanism: its metric(s). There is a wide disparity in terms of both the number and the type of metrics used.

Quantity of Metrics

Simple routing protocols support as few as one or two routing metrics. More sophisticated protocols can support five or more metrics. It is safe to assume that the more metrics there are, the more varied and specific they are. Therefore, the greater the variety of available metrics, the greater your ability to tailor the network's operation to your particular needs. For example, the simple distance-vector protocols use a euphemistic metric: distance. In reality, that distance is not related at all to geographic mileage, much less to the physical cable mileage that separates source and destination machines. Instead, it usually just counts the number of hops between those two points.

Link-state protocols may afford the capability to calculate routes based on several factors:

Most of these factors are highly dynamic in a network; they vary by time of day, day of week, and so forth. The important thing to remember is that as they vary, so does the network's performance. Therefore, the intent of dynamic routing metrics is to allow optimal routing decisions to be made using the most current information available.

Static Versus Dynamic Metrics

Some metrics are simplistic and static, whereas others are highly sophisticated and dynamic. Static metrics usually offer the capability to customize their values when they are configured. After this is done, each value remains a constant until it is manually changed.

Dynamic protocols enable routing decisions to be made based on real-time information about the state of the network. These protocols are supported only by the more sophisticated link-state or hybridized routing protocols.

Summary

One of the most important decisions in the development of an internetwork is the selection of a routing protocol or protocols. Such selection should be done carefully and with an appreciation for the long-term implications of your decisions. This chapter provides a generic overview of the various ways that routing can be performed, as well as the benefits and limitations of each. Part III builds on this overview with specific details about how the more commonly encountered routing protocols operate.


hometocprevnextglossaryfeedbacksearchhelp
Posted: Wed Mar 3 16:19:54 PST 1999
Copyright 1989-1999©Cisco Systems Inc.
Copyright © 1997 Macmillan Publishing USA, a Simon & Schuster Company