IP Routing Protocols

IP supports a broad variety of IGPs. In the distance vector category, IP supports the Routing Information Protocol (RIP) and the Interior Gateway Routing Protocol (IGRP). In the hybrid category, IP supports the Enhanced Interior Gateway Routing Protocol (EIGRP). In the link-state category, IP supports the Open Shortest Path First (OSPF) protocol and the Integrated Intermediate System to Intermediate System (Integrated IS-IS) protocol. IP also supports two EGPs: the Exterior Gateway Protocol (EGP) and the Border Gateway Protocol (BGP).

RIP is the original distance vector protocol. RIP and its successor, RIP version 2 (RIPv2), enjoyed widespread use for many years. Today, RIP and RIPv2 are mostly historical. RIP employs classful routing based on classful IP addresses. RIP distributes routing updates via broadcast. RIPv2 enhances RIP by supporting classless routing based on variable-length subnet masking (VLSM) methodologies. Other enhancements include the use of multicast for routing update distribution and support for route update authentication. Both RIP and RIPv2 use hop count as the routing metric and support load balancing across equal-cost paths. RIP and RIPv2 are both IETF standards. For more information about classful/classless routing and VLSM, see Chapter 6, "The OSI Network Layer."

IGRP is a Cisco Systems proprietary protocol. IGRP was developed to overcome the limitations of RIP. The most notable improvement is IGRP's use of a composite metric that considers the delay, bandwidth, reliability, and load characteristics of each link. Additionally, IGRP expands the maximum network diameter to 255 hops versus the 15-hop maximum supported by RIP and RIPv2. IGRP also supports load balancing across unequal-cost paths. IGRP is mostly historical today.

EIGRP is another Cisco Systems proprietary protocol. EIGRP significantly enhances IGRP. Although EIGRP is often called a hybrid protocol, it advertises routing-table entries to adjacent routers just like distance vector protocols. However, EIGRP supports several features that differ from typical distance vector protocols. Among these are partial table updates (as opposed to full table updates), change triggered updates (as opposed to periodic updates), scope sensitive updates sent only to affected neighbor routers (as opposed to blind updates sent to all neighbor routers), a "diffusing computation" system that spreads the route calculation burden across multiple routers, and support for bandwidth throttling to control protocol overhead on low-bandwidth WAN links. EIGRP is a classless protocol that supports route summaries for address aggregation, load balancing across unequal-cost paths, and route update authentication. Though waning in popularity, EIGRP is still in use today.

OSPF is another IETF standard protocol. OSPF was originally developed to overcome the limitations of RIP. OSPF is a classless protocol that employs Dijkstra's Shortest Path First (SPF) algorithm, supports equal-cost load balancing, supports route summaries for address aggregation, and supports authentication. To promote scalability, OSPF supports the notion of areas. An OSPF area is a collection of OSPF routers that exchange LSAs. In other words, LSA flooding does not traverse area boundaries. This reduces the number of LSAs that each router must process and reduces the size of each router's link-state database. One area is designated as the backbone area through which all inter-area communication flows. Each area has one or more Area Border Routers (ABRs) that connect the area to the backbone area. Thus, OSPF implements a two-level hierarchical topology. All inter-area routes are calculated using a distance-vector algorithm. Despite this fact, OSPF is not widely considered to be a hybrid protocol. OSPF is very robust and is in widespread use today.

IS-IS was originally developed by Digital Equipment Corporation (DEC). IS-IS was later adopted by the ISO as the routing protocol for its Connectionless Network Protocol (CLNP). At one time, many people believed that CLNP eventually would replace IP. So, an enhanced version of IS-IS was developed to support CLNP and IP simultaneously.

The enhanced version is called Integrated IS-IS. In the end, the IETF adopted OSPF as its official IGP. OSPF and Integrated IS-IS have many common features. Like OSPF, Integrated IS-IS is a classless protocol that employs Dijkstra's SPF algorithm, supports equal-cost load balancing, supports route summaries for address aggregation, supports authentication, and supports a two-level hierarchical topology. Some key differences also exist. For example, Integrated IS-IS uses the Dijkstra algorithm to compute inter-area routes.

EGP was the first exterior protocol. Due to EGP's many limitations, many people consider EGP to be a reachability protocol rather than a full routing protocol. EGP is mostly historical today. From EGP evolved BGP. BGP has since evolved from its first implementation into BGP version 4 (BGP-4). BGP-4 is widely used today. Many companies run BGP-4 on their Autonomous System Border Routers (ASBRs) for connectivity to the Internet. Likewise, many ISPs run BGP-4 on their ASBRs to communicate with other ISPs. Whereas BGP-4 is widely considered to be a hybrid protocol, BGP-4 advertises routing table entries to other BGP-4 routers just like distance vector protocols. However, a BGP-4 route is the list of AS numbers (called the AS_Path) that must be traversed to reach a given destination. Thus, BGP-4 is called a path vector protocol. Also, BGP-4 runs over TCP. Each BGP-4 router establishes a TCP connection to another BGP-4 router (called a BGP-4 peer) based on routing policies that are administratively configured. Using TCP relaxes the requirement for BGP-4 peers to be topologically adjacent. Connectivity between BGP-4 peers often spans an entire AS that runs its own IGP internally. A TCP packet originated by a BGP-4 router is routed to the BGP-4 peer just like any other unicast packet. BGP-4 is considered a policy-based routing protocol because the protocol behavior can be fully controlled via administrative policies. BGP-4 is a classless protocol that supports equal-cost load balancing and authentication.

Note

BGP-4's use of TCP can be confusing. How can a routing protocol operate at OSI Layer 3 and use TCP to communicate? The answer is simple. For a router to operate, it must gather information from peer routers. Various mechanisms exist for gathering such information. Once the information is gathered, the subsequent functions of path determination and packet forwarding are executed at OSI Layer 3. In the case of BGP-4, the peer communication function leverages TCP, but AS_Path creation and packet forwarding are executed at OSI Layer 3.

The sheer volume of information associated with IP routing protocols can be very intimidating to someone who is new to IP networking. For more information on IP routing protocols, readers can consult the numerous IETF RFCs in which the protocols are defined and enhanced. Alternately, readers can consult one of the many books written about this subject. A very comprehensive analysis of all IP routing protocols is available in the two-volume set by Jeff Doyle entitled Routing TCP/IP, volumes I and II.