Classification of IP Routing Protocols

Routing between the various segments of a network can be achieved by programming the routers with manual routing information, commonly referred to as static routes, or by using a dynamic protocol to automate the collection of routing information and intelligence. The applications or protocols used in the latter case are referred to as dynamic routing protocols. In addition, a router can just use a default route for forwarding packets heading to nondirectly connected subnets. The next hop along the default route is referred to as the default gateway. Default routes can be generated by either static or dynamic methods . Some common dynamic IP routing protocols in use today are the Routing Information Protocol (RIP) version 1 and version 2, Interior Gateway Routing (IGRP), Enhanced Interior Gateway Protocol (EIGRP), Integrated Intermediate System-to-Intermediate System Routing Protocol (IS-IS), Open Shortest Path Routing Protocol (OSPF), and the Border Gateway Protocol (BGP).

The different breeds of routing protocols have different capabilities related to both architectural design and embedded functionality. The following sections differentiate between commonly used routing protocols. The material covered focuses mostly on unicast routing protocols and distinctions are primarily provided for such protocols.

The following classifications are covered:

• Classful versus classless protocols

• Intradomain versus interdomain protocols

• Distance-vector versus link-state protocols

Classful Versus Classless Protocols

Dynamic IP routing protocols can be categorized into classful and classless protocols, as shown in Table 1-2.

Table 1-2. Classful and Classless Protocols

Classful Routing Protocols	Classless Routing Protocol
RIP version 1	RIP version 2
IGRP	EIGRP
	Integrated IS-IS
	OSPF
	BGP

Classful protocols obtain routing information through the exchange of network numbers without mask information and are designed to deal with only classful IP addresses. When IP subnetworks are exchanged, classful protocols arbitrarily apply known masks for matching classful entries that are locally configured. Therefore, such protocols are confused by the notion of a variable prefix length embodied in CIDR. However, classless protocols exchange masking information in addition to network numbers when collecting routing information, allowing them to work flexibly with VLSMs or classless addressing. Table 1-3 summarizes the differences between classful and classless protocols.

Table 1-3. Attributes of Classful and Classless Protocols

Classful Routing Protocols	Classless Routing Protocols
Do not support supernets	Support supernets
Do not carry subnet masks in updates	Advertise routes together with subnet masks
Cannot handle VLSMs	Support VLSMs
Do not support discontiguous subnets	Support discontiguous subnets
Mostly periodic updates	Mostly incremental updates
Simple to configure and troubleshoot on Cisco routers	Simple to complex to configure on Cisco routers, and complex to troubleshoot

Intradomain Versus Interdomain Routing

A network of interconnected routers and related systems managed and maintained together by a common administration is often referred to as a network domain. A network domain is also sometimes called an autonomous system. The Internet consists of several interconnected network domains spanning the whole world. Routing protocols are designed and optimized for use within a domain (intradomain) or between domains (interdomain). Figure 1-7 shows three network domains (AS1, AS2, and AS3) interconnected into a global routing system. Also depicted in the diagram are instances of intradomain and interdomain routing.

Figure 1-7. Intradomain and interdomain routing.

All the routing protocols listed in the preceding section, with the exception of BGP, are optimized for intradomain functionality. Intradomain routing protocols typically do not offer flexibility for policy implementation and also cannot deal with a large number of routes, such as in the global Internet routing table. Obviously, more complex policies will be required to control the exchange of routing information between two separate domains, and this is certainly what BGP is optimized for.

Because of the vastness of the global Internet, the number of routes that need to be handled globally is large. The size of the global Internet routing table is currently in the order of 100,000.00 routes and is slated to grow further. Therefore, an interdomain routing protocol must provide the following basic capabilities:

• Support configuration of complex policies, such as route filtering, tagging, and so on

• Handle a large number of routes under both stable and unstable conditions

• Respond reasonably fast to network changes ”that is, sending updates, receiving and processing updates, and selecting alternative routes

An interdomain protocol also must handle multiple peers, with each peer presenting a different view of the same large tables. Currently, only BGP can deliver on all these requirements, and it remains the de facto interdomain routing protocol on the Internet. The current version of BGP for unicast IP routing, version 4, is specified by RFC 1771. BGP is still evolving and has had many enhancements in recent times, including the addition of multiprotocol capabilities in RFC 2858. RFC 2858 specifies multiprotocol extensions to BGP4, providing the architectural framework for multiprotocol label-switching virtual private networks (MPLS VPNs), interdomain multicast routing, and support for IPv6 interdomain routing.

Distance-Vector Versus Link-State Protocols

This chapter discusses the classification of routing protocols based on architectural design (distance-vector and link-state protocols). Essentially, this categorization applies to intradomain protocols, also known as Interior Gateway Protocols (IGPs). The only surviving interdomain protocol, BGP, was discussed in the preceding section. BGP is normally considered a path-vector protocol because a key attribute it ascribes to routes is a vector of path information known as AS Path. An AS Path is a vector of autonomous systems that a route has traversed.

Table 1-4 shows the classification of IGPs into distance-vector and link-state protocols. IGPs are easily placed into these two categories. Distance-vector protocols are simpler in design and tend to be classful. This classful attribute disqualifies them from being viable options for routing in most large modern networks. Distance-vector protocols use periodic update mechanisms that consume a lot of bandwidth resources. Link-state protocols send only incremental updates for any network changes. In general, link-state protocols require a lot more processing and memory resources than distant -vector protocols. They don't have some of the inherent problems associated with distance-vector protocols, however, such as periodic updates, transient loops , count to infinity, and slow convergence issues.

Link-state protocols are more processor- intensive than distance-vector protocols. This is true because a change in the topology of an area would normally trigger a complete SPF run over the entire link-state database, whereas distance-vector protocols perform their computation on the basis of individual routes, requiring less computation for minor topology changes.

A critical advantage of link-state protocols over distance-vector protocols is the capability to support hierarchical network architectures, giving them high potential to scale. Distance-vector protocols work in only flat network architectures and generally have limited scaling capabilities.

Table 1-4. Distance-Vector and Link-State Protocols

Protocol	Category	Metric	Algorithm
RIP v1	Distance Vector	Hop count	Bellman-Ford
RIP v2	Distance Vector	Hop count	Bellman-Ford
IGRP	Distance Vector	Composite	Bellman-Ford
EIGRP	Distance Vector	Composite	Diffusing Update Algorithm
OSPF	Link State	Bandwidth-based cost	Shortest Path First
IS-IS	Link State	Manual cost	Shortest Path First

IGRP and EIGRP

IGRP and EIGRP are proprietary protocols developed by Cisco Systems. As indicated by their names , EIGRP is based on IGRP and features critical enhancements, such as support for VLSM, classless routing. EIGRP uses a different algorithm for route computation, as indicated in Table 1-4. Another improvement over IGRP is incremental updates and faster convergence using the concept of feasible successor. EIGRP also features multiprotocol capabilities, supporting routing for IP, Internet Packet Exchange Protocol (IPX), and AppleTalk, whereas IGRP is limited to only IP.

RIP v1 and RIP v2

RIP version 1 and version 2 have origins in the IETF. Version 2 introduces enhancements, such as multicast updates and advertisement of routes, with mask information to support VLSMs and classless addressing.

OSPF

OSPF also originated from the IETF and is probably the most popular and well- understood IGP because of its originality as an IP protocol and extensive coverage in internetworking literature. OSPF has evolved over time into a robust protocol, acquiring the necessary capabilities to build complex routing infrastructures in both enterprise and service provider environments. Although the entirety of the OSPF protocol is complex, its basic concepts and capabilities are not difficult to understand and configure on Cisco routers.

Integrated IS-IS

This book is about the IS-IS protocol and the following chapters provide insights into this protocol's innards and capabilities. IS-IS (ISO 10589) was originally specified by the International Telecommunications Union (ITU), formally the International Organization for Standardization (ISO), as a routing protocol for connectionless network layer protocol (CLNP). IS-IS was first implemented for routing within the DECnet Phase V architecture at Digital Equipment Corporation (DEC). It was later adapted to support IP routing by the IETF in RFC 1195. Despite their different functional designs and origins, IS-IS and OSPF share a lot in common. Both are link-state protocols and use the SPF algorithm for route computation.