The matrix as shown in Table 3-2 provides a list of the more important features to consider when weighing the technical issues concerning the selection of a routing protocol.
All routing protocols have three important characteristics when dealing with the issue of convergence:
RIP, IS-IS, and OSPF detect certain types of changes instantly. In general, any change that can be detected by a physical change (such as loss of carrier) will be detected immediately by any routing protocol.
In addition, all three routing protocols have mechanisms to detect other failures (such as the loss of an adjacent router or the degradation of an interface to the point where it should no longer be used). All three protocols cause adjacent routers to exchange information periodically.
After the routing protocol has detected the topology change, it needs to adjust the routing tables to accommodate the new topology. RIP generally updates its routing tables every 30 seconds (this is the default time), although during the normal update process, RIP sends out its entire routing table. OSPF and Integrated IS-IS have two mechanisms for updating routing tables. If the topology change were within the area, all of the existing routes affected by the change would be discarded and a new routing table would be generated. In general, OSPF and Integrated IS-IS converge in less than two seconds. The amount of CPU required to do the recompilation is strongly affected by the number of routes and the amount of redundancy in the network.
All routing protocols exchange routing information dynamically. The three most important and interesting questions concerning the operation of routing updates are as follows:
Variable-Length Subnet Masks
RIPv2, OSPF, and Integrated IS-IS all include support for variable-length subnet masks (VLSM). VLSM is required to support route summarization. In addition, VLSM also enables network administrators to use their address space more effectively.
Todays networks are commonly designed with redundant paths. This has two positive benefits: re-routing in case of failure and load sharing. All routing protocols supported by Cisco provide load sharing across up to four equal-cost paths.
The quality of route selection is essentially controlled by the value of the metrics placed upon the various routes. There are two components of importance in how a routing protocol uses metrics: the range of the values the metric can express and how the metric is computed.
RIP uses a very simple hop count metric that can be expressed with values between 0 and 16. The computation is also very straightforward in that the cost determined for the metric is a matter of how many hops the destination is from the source router.
OSPF uses a flat metric with a range of 16 bits. This results in OSPF having a metric range that is from zero to 65,535. By default, OSPF metrics are assigned as an inverse of the bandwidth available on an interfacenormalized to give FDDI a metric of 1. OSPF computes the cost of a path by summing the metrics for each hop on that path.
Integrated IS-IS uses a flat metric. The metric range is 0-1,023. By default, all Integrated IS-IS metrics are 10. Network administrators need to configure non-default values. Integrated IS-IS computes the cost of a path by summing the metrics for each hop on that path.
RIP has many problems associated with scaling into the larger network sizes. For example, if a network has more than 15 hops, RIP will run into problems due to its design limitations.
As for IS-IS, ISO 10589 states that 100 routers per area and 400 L2 routers should be possible. The biggest scaling issue now seems to be the overhead of the flooding in large meshed networksfor example, flat ATM clouds with many routers attached that form a full mesh.
OSPF, on the other hand, scales very well no matter how large the network is. In order to make the network operate optimally, however, you should implement physical and logical areas, as needed.
Table 3-3 documents some of the more important features to consider when weighing the business issues concerning the selection of a routing protocol.