RIP Routing Convergence

Any time there is a change in the network topology, each router must converge on the change the most significant of which is a change to an immediate neighboring router. Simply stated, convergence is the mechanism by which each router agrees with the other on what the new network topology looks like.

In the previous figure, if the link between Router C and Router D fails, each interconnected router will converge and update its routing tables as a consequence. If Router B fails, the same process occurs whereby each Router A, C, D, and E will update its tables, marking any path involving Router B as being unavailable.

There are several mechanisms in place that affect the convergence of RIP routers:

• Count to Infinity ()

• Poison Reverse

• Triggered Updates

• Hold-Down Timers

Each of these is discussed in the following sections.

Count to Infinity ()

Counting to infinity is when each router adds 1 to the hop count before advertising the route. This hop count continues to increase by 1 until infinity is reached, rendering the destination network unreachable. In RIP implementations, infinity is 16 hops. Figure B-2 illustrates a three-node network that has experienced a link failure between Router A and Router C.

Figure B-2. RIP Network with Failed Link

Table B-3 shows the routing table for each router prior to the network failure.

Table B-3. Routing Table Prior to Failure

Router Name	Destination Host	Next Hop	Number of Hops	Via Network Path
A	Any B Network	B	1	Directly Connected
		C	2	C-B
	Any C Network	C	1	Directly Connected
		B	2	B-C
B	Any A Network	A	1	Directly Connected
		C	2	C-A
B	Any C Network	C	1	Directly Connected
		A	2	A-C
C	Any A Network	A	1	Directly Connected
		B	2	B-A
	Any B Network	B	1	Directly Connected
		A	2	A-B

Something unique starts to happen when Routers A and C detect the failed link, however. Router A is trying to connect to Router C, but it has no direct connection. Router A learns that Router B has a connection to Router C. However, Router B advertises that it also can get to Router C directly or through Router A; Router A in turn advertises that it can get to Router C through Router B. This will go back and forth between the two routes, adding 1 to the hop count, until the next routing update (180 seconds) takes place.

Table B-4 shows what the routing table will look like after each router has counted to infinity to determine the reachability of each node after the network link failure.

Table B-4. Routing Table After Link Failure

Router Name	Destination Host	Next Hop	Number of Hops	Via Network Path
A	Any B Network	B	1	Directly Connected
		C	2	C-B
	Any C Network	C	1	Directly Connected
		B	2	B-C
B	Any A Network	A	1	Directly Connected
		C	16	Unreachable
	Any C Network	C	1	Directly Connected
		A	16	Unreachable
C	Any A Network	A	16	Unreachable
		B	2	B-A
	Any B Network	B	1	Directly Connected
		A	2	A-B

The issue here is the amount of time taken for the 16 hop count (unreachable) to be achieved. During this time, datagram traffic is circling around between the two nodes, never reaching its ultimate destination until the next routing update is converged upon.

There are two methods used to avoid the count-to-infinity problem: split horizon and triggered updates. These are discussed in the next sections.

Split Horizon

Split horizon essentially divides the routed network (the horizon) into logical pieces. Split horizon is based on a simple premise: The router will not advertise a route over the same interface from which it was learned.

NOTE

In a Frame Relay network where the network manager/administrator has implemented multiple subinterfaces, the recommendation is to disable split horizon on the serial interface if you want remote sites to see each other across the network.

There is a drawback to simply implementing split horizon each router must wait for the destination to be marked as unreachable. By the time a route has timed out and been flushed from the table (a process that takes six update messages, at 30 seconds each), more than three minutes have passed before each routing table is updated with the inactive link. During this time, there are five update intervals that can pass where each router can misinform another as to the reachability of certain destinations. Split horizon coupled with a poison reverse addresses and solves this problem.

RIP Split Horizon with Poison Reverse

Where split horizon is designed to prevent routing loops in an internetwork, split horizon with poison reverse makes this a bit more effective in that six update cycles do not have to pass to stop a routing loop (see the sections, "Count to Infinity" and "Split Horizon," earlier in this chapter). Split horizon with poison reverse takes a more proactive stance in managing and updating the routing tables in that upon detection of an inactive link, RIP with poison reverse sets the metric for that destination to infinity for the next routing update.

Although split horizon with poison reverse is the preference over (standalone) split horizon, there are still concerns with larger internetworks with multiple paths in that RIP is still subject to the counting to infinity problem of routing updates. Triggered updates were introduced to solve the problem of routing loops caused by the "counting to infinity" operations.

RIP Triggered Updates

Triggered updates are used to speed up convergence of a RIP routed network. Triggered updates are rules in the routing protocol that require routers to immediately broadcast an update message whenever there is a change to a route metric, without waiting for the next 30-second regular update interval to pass.

Triggered updates are designed to overcome the time issues that are still involved when dealing with split horizon or split horizon with poison reverse.

RIP Hold-Down Timers

Although triggered updates are a significant mechanism compared to split horizon and poison reverse, there is still the issue of time. Will each router in the internetwork receive and update its tables in a reasonable amount of time, an interval that passes before traffic is to be transmitted?

Hold-down timers solve this potential problem by working in conjunction with triggered updates. Essentially, when a triggered update has been sent, a clock starts counting down (to zero). Until this hold-down timer hits zero, the router will not accept any neighbor updates for the route in question.

The use of a hold-down timer prevents a RIP router from accepting and converging on updates for a route that has been invalidated over a period of time. Hold-down timers prevent a router from believing that another router may have a path to an invalid destination.