|‚ < ‚ Free Open Study ‚ > ‚|
Static and Dynamic Routes
Contrast this behavior with static routes, which are manual route entries and require manual intervention to
Even though static routing is less demanding on system resources and requires a lower level of technical skill to configure and troubleshoot, the sheer effort of manually entering routes for a sizeable network makes it a less attractive option. Obviously, static routing is not a good candidate for today's large enterprise and Internet service provider (ISP) IP-based networks. Another drawback to static routing is that it is less flexible for implementation of complicated routing policies. When it comes to routing policy implementation, there is no better substitute for the intelligence and flexibility provided by dynamic routing protocols, such as BGP, OSPF, and IS-IS. The next section further discusses dynamic routing protocols.
|‚ < ‚ Free Open Study ‚ > ‚|
|‚ < ‚ Free Open Study ‚ > ‚|
The last section discusses the essence of IP routing and indicates that dynamic automatic routing is very necessary for large network deployments. This section discusses the characteristics and classification of various IP routing protocols. Although all routing protocols have a common goal of gathering routing information to support packet-forwarding decisions, they can be
The previous section indicates that IP provides an addressing scheme for identifying various locations or subnets in the network. The destination IP address in an IP packet indicates the target address of the packet. The sender's address is stored in the source address field. An important concept to understand about IP addressing is IP subnetworks. IP subnetworks ‚ or subnets, for short ‚ are mentioned earlier in the section on IP address-ing concepts. Physically, an IP subnet is a collection of
The earlier section "IPv4 Address Classes" discusses unicast and multicast addresses. The unicast address space is used for addressing network devices, whereas addresses from the multicast space are used for specifying groups or users
For any IP unicast subnet, the last address, such as in 192.168.1.255/24, is known as the broadcast address. This address can be used to target all nodes on the subnet at the same time in what is referred to as a directed broadcast.
A unicast routing protocol is optimized for processing unicast network information and provides routing intelligence for forwarding IP packets to unicast destination addresses. Multicast forwarding is conceptually different and requires special routing applications to support forwarding of multicast packets.
Unicast Versus Multicast IP Routing
Two devices in an IP network normally communicate by sending unicast traffic to each other's IP address. An IP node might have many active interfaces, each of which needs to be configured with an IP address from the unicast space. The address on an interface uniquely defines the device on the subnet directly connected to that interface.
Cisco routers also support the concept of secondary logical subnets, many of which can be configured on a router's interface in addition to the primary address on that interface. Additionally, you can enable tunnel and loopback interfaces on a Cisco router, both of which provide it with unicast IP reachability. Packets with unicast addresses in their destination field are forwarded based on information in the IP routing table. The IP routing table on a Cisco router is displayed with the show ip route command.
If the address in the destination field of a packet is from the multicast address space (Class D), the packet is directed to a multicast
Figure 1-7 illustrates a situation in which a packet is forwarded from SRC1 to two separate destinations, RCV1 and RCV2, by unicast forwarding.
Figure 1-7. Multidestination Unicast Forwarding
In this case, SRC1 generates two identical streams of packets with destination addresses 10.1.1.1 and 10.1.1.2, respectively. Packets
Figure 1-8. Multicast Forwarding
Multicast forwarding provides a more efficient way to deliver information by replicating packets only at fork points of the network where paths to receivers follow
Multicast routing protocols are functionally different from unicast routing protocols, in that they build multicast forwarding state in the multicast-enabled routers by using a concept known as
RPF is discussed further in Chapter 12, "Understanding Protocol Independent Multicast (PIM)," which covers IP Multicast routing.
Table 1-2 shows a table of popular multicast and unicast routing protocols.
Table 1-2. Unicast and Multicast Routing Protocols
All the listed unicast routing protocols are supported in Cisco IOS Software; however, from the listed multicast routing protocols, only Protocol Independent Multicast (PIM) sparse mode/dense mode (SM/DM), Multicast Source Discovery Protocol (MSDP), and Multiprotocol BGP are supported.
Multicast routing environments also need an additional protocol called the Internet Gateway Multicast Protocol (IGMP). Multicast OSPR (MOSPF) is not supported at all, but IOS provides special capabilities for interoperability with the Distance Vector Multicast Routing Protocol (DVMRP).
As of this writing, multicast routing protocols are not widely deployed on the Internet. However, this situation obviously will change in the near future as more multicast-oriented applications, such as radio broadcasting, video streaming, remote training, videoconferencing, and gaming, become more popular on the Internet.
Classless Versus Classful IP Routing Protocols
The concepts of classless and classful IP routing protocols have roots in the manner in which IP addresses originally were defined.
Under classful addressing rules, a network number was assumed to retain its natural mask unless explicitly specified when
The significant growth of the Internet to global dimensions called for more efficient use of the limited IPv4 address space. Available addresses in the IP address space therefore attained the status of a
Table 1-3 features a list of routing protocols categorized as classful and classless. RIP-1 and IGRP are grouped under classful protocols, whereas the more recently developed RIP-2, EIGRP, OSPF, IS-IS, and BGP fall in the classless category. The Exterior Gateway Protocol (EGP), the predecessor of the Border Gateway Protocol (BGP), which currently is
Table 1-3. Classful and Classless IP Routing Protocols
Interior Gateway Protocols Versus Exterior Gateway Protocols
Even though many unicast routing protocols were developed in the early days of the ARPANET (the predecessor to the Internet), Routing Information Protocol (RIP) emerged as the most popular. Many independent networks that were created at
Figure 1-9 shows two routing domains, AS 65001 and AS 65002, and an overlapping (shaded) region depicting the interconnection between border routers from each domain. In more current routing terminology, a routing domain also is referred to as an autonomous system . An autonomous system is an independent routing domain under the control of a single administrative authority.
Figure 1-9. Intradomain and Interdomain Routing
As noted before, an exterior gateway protocol provides the capability for sharing routing information between the two domains. Currently at version 4, BGP is the only IP
Table 1-4. IGP and EGP Classification
The Interior Gateway Routing Protocol (IGRP) was invented by Cisco Systems to offer better metrics than the simple hop count supported by RIP. IGRP introduced a composite metric that consists of several parameters:
Cisco evolved IGRP into the Enhanced Interior Gateway Routing Protocol (EIGRP). EIGRP provides faster convergence relative to IGRP by using backup routes, referred to as
OSPF and IS-IS are both popular IGPs used in very large IP networks. IS-IS originally was designed as a routing protocol for the Connectionless Network Protocol (CLNP) but later was
Also, OSPF and IS-IS are link-state protocols that use the shortest path first (SPF) algorithm (named after Dijkstra) for route computation, making them converge relatively fast in re-sponse to network changes.
Both protocols also support a
An interesting point to note, however, is that OSPF was designed entirely for IP only, and OSPF packets are encapsulated in IP packets. In contrast, IS-IS was designed for CLNP and was adapted to support IP additionally. IS-IS packets are not encapsulated in IP packets but rather directly by the data link protocol.
Distance Vector Versus Link-State Protocols
This section takes a look at routing protocol from a different perspective. In the previous sections, we considered general classification such as classful versus classless and also IGP versus EGP. This section discusses classification based on design and operation. The second row in Table 1-4 places the protocols discussed so far into four different categories, two of which stand out ‚ distance vector and link-state. These two broad categories actually apply to IGP as shown in the table.
Versions 1 and 2 of RIP (RIP-1 and RIP-2) and IGRP are classified as distance vector protocols because they use route-computation algorithms based on the Bellman-Ford algorithm. The Bellman-Ford algorithm is used in graph theory for calculating the shortest distance between two vertices in a directed graph. A directed graph is a collection of points, interconnected with directional links, such as the nodes and links in an internetwork. Routers running distance vector routing protocols use the Bellman-Ford algorithm for determining the shortest paths to all known locations in the network.
OSPF and Integrated IS-IS are both link-state protocols and use the shortest path first algorithm (Dijkstra) for route computation. Just like the Bellman-Ford algorithm, the Dijkstra algorithm provides an alternate method for computing the shortest distance between two points in a directed graph.
EIGRP uses a Cisco Systems ‚ patented algorithm known as the Diffusing Update Algorithm (DUAL) to optimize route calculation, breaking away from its predecessor, IGRP, which is based on the Bellman-Ford algorithm.
The type of algorithm used by a protocol for route computation goes a long way toward
Distance Vector Routing Concepts
This section reviews key concepts that underlie the operation of distance vector routing protocols, such as metrics, count to infinity, split horizon, holddowns, and triggered updates. These concepts are evaluated in terms of general routing functionality, such as stability and speed of convergence and loop
Distance Vector Metrics
In the Bellman-Ford algorithm, each router advertises the best paths to all known des-tinations, from its perspective, to all neighbors. The links between routers are assigned a measure known as cost or metric. The metric can be determined from characteristics of the links, such as hop count, bandwidth, delay, reliability,
RIP uses hop count for metric, with the maximum possible number of hops to any
IGRP, which is also a distance vector protocol, uses a metric system that takes into consider-ation relevant characteristics of the network, such as bandwidth, associated maximum trans-mission unit, reliability of links, and also path delay. The metric assigned to each link in the outgoing direction is calculated from a formula that takes into consideration all these char-acteristics. This
The Bellman-Ford algorithm uses a vector (distance vector), consisting of cost (metric) and next-hop information for each known route to determine best paths in the network from any standpoint. An iterative procedure calculates the cost of all paths for any received route and selects the vector with the best cost for each route. Hence, routing protocols that are based on the Bellman-Ford algorithm commonly are referred to as distance vector protocols (see Table 1-4).
When there is a topology change, a router might invalidate some of the previously known best paths. The router then uses new or existing information to determine an alternate best path for each affected destination. Recalculating routes to rediscover alternate routes as a result of network topology changes is referred to as routing convergence. Routing convergence may be triggered by events such as router failures, link failures, or even administrative metric adjustments.
Distance vector protocols such as RIP and IGRP are relatively simple compared to their link-state counterparts. However, this simplicity comes with a price. Because each router bases its best-path determination on the best paths advertised by neighbors, such protocols are very prone to routing loops. A routing loop occurs when two nodes point to each other as the next hop along the path to the same destination. The most obvious effect of routing loops is that they prolong the time it takes for a router to determine a route is no longer available or to select an alternate path. Routing loops adversely impact convergence times. Therefore, it is desirable that unusable routes be removed from the network as soon as possible. The following sections discuss various
Routers running distance vector protocols determine best paths for routes relative to neigh-bors that have advertised those routes to them. The mechanics of operation of distance vector protocols,
Figure 1-10. Routing Loops in Distance Vector Environments
In Figure 1-10, RT1, RT2, RT3 are connected
Counting to Infinity
To prevent routing loops of indefinite duration, distance vector protocols enforce limits on route metrics that allows routers to declare routes as unreachable after the associated metrics reach a certain value. In the loop situation described in Figure 1-10, RT1 and RT4 might advertise Dest3 to each other, each time increasing the associated hop count received from the other by 1 and before readvertising the route. Consequently, the metric associated with Dest3 will continue to increase. Counting to infinity places an upper bound on the metric beyond which it is considered infinity and the route is declared unusable. For RIP, this upper bound is 15.
Holddown is used to dampen a route's response action to finding an alternate route when a primary route is no longer usable. When a router determines that a route is no longer avail-able, it places the route in holddown state for a duration called the holddown time, during which it doesn't select an alternate route, even if available. The route in holddown state is advertised with a metric or value of infinity in an attempt to purge it from the network. Purging unusable routes helps reduce the incidence of routing loops.
Another benefit to using holddowns is that it
Split Horizon and Poison Reverse
Routing loops are primarily the result of routes being leaked back to their sources. For example, in Figure 1-10, the loop between RT1 and RT2 is caused by feedback of Dest3 back to RT2 by RT1, misleading RT2 to think that RT1 is the next hop on an alternate path to Dest3.
Split horizon prevents a router from advertising a route back out the interface through which it was received. With split horizon in effect, RT1 cannot advertise Dest3 back to RT3 over the link between them (see Figure 1-10).
Poison reverse is similar in principle to split horizon, except that it allows routes to be advertised back out the interfaces on which they were received as unreachable (metric of infinity assigned). That is, routes are "poisoned" in the reverse direction. Referring to Figure 1-10, with poison reverse enabled, RT1 advertises Dest3 back to RT2, but with a metric value of infinity (16 hops, in the case of RIP).
The approach adopted by poison reverse can result in undue waste of bandwidth if many poisoned routers must be advertised back out. However, this approach speeds up route convergence by eliminating the need for holddowns. In this case, the alternate route would have an obvious infinite metric when fed back to the source, hence simplifying the search for an alternative path, when the primary route is lost.
Periodic and Triggered Updates
Routers running distance vector routing protocols, such as RIP and IGRP, advertise all the contents of their routing tables at regular intervals. Periodic broadcasts of large routing tables are a major concern in large networks. For example, RIP broadcasts all known routes out of every active interface every 30 seconds, by default, even if there are no changes. IGRP uses a default update interval of 90 seconds.
If updates are advertised only periodically, changes in the network might not be communi-cated fast enough,
Triggered (or flash) updates remove delays in convergence caused by periodic updates by sending updates immediately following a network change instead of waiting for the periodic update timer. Flash updates trickle through the network from one node to the other, resulting in an overall time gain in network-wide convergence, even if not very significant. Complicity between periodically scheduled updates and triggered changes can result in unpredictable behavior.
Link-state protocols are relatively more modern and, therefore,
Unlike distance vector protocols, which share best-known routing information, link-state protocols allow routers to exchange topology (link-state) information that allows them to draw out the layout of the internetwork's topology. Routers in a link-state network converge relatively faster than their distance vector counterparts by responding immediately to changes in the topology, without the need for loop avoiding or limiting holddowns and counting to infinity. For example, RIP and IGRP typically feature convergence times in minutes, whereas OSPF and IS-IS converge in the order of seconds for comparable network changes.
Link-state protocols support hierarchy for scaling purposes by carving out a network into areas (see Figure 1-11). Routing within areas fall in the first level of the routing hierarchy. The areas are interconnected over a backbone area, and routing within the backbone consti-tutes the second level of the hierarchy.
Figure 1-11. Areas and Hierarchy in Link-State Protocols
Routers in the same area or the backbone share link-state information that is
Metrics in Link-State Protocols
Both OSPF and IS-IS use metrics, which are measures of link bandwidth. OSPF goes a step further, to provide autoconversion of the bandwidth on interface to a link cost. IS-IS metrics are 10, by default, on all interfaces. In both cases, the metric or cost associated with a link can be manually configured. The metric associated with a route is the sum of all the metrics on the outgoing links to the associated destination.
Chapters 8 and 10 provide more information on metrics in OSPF and IS-IS, respectively.
|‚ < ‚ Free Open Study ‚ > ‚|