Communications Server for z/OS V1R7 TCP/IP Implementation, Volume 1: Base Functions, Connectivity, and Routing

Chapter 5. Routing 153

Subnet loses its importance, replaced by Link (since multiple IPv6 prefixes per link are

allowed and expected, routing by subnet/prefix makes less sense). In OSPFv4, most routing

is done by subnet. In OSPFv3 it is done by link. This is because in IPv6 a

subnet (prefix) does

not always uniquely identify a link, and a link can have more than one prefix assigned.

5.1.5 Choosing the routing protocol

The choice of a routing protocol is a major decision for the network administrator. It has a

major impact to overall network performance. The selection depends on network complexity,

size, and administrative policies. The protocol chosen for one type of network may not be

appropriate for other types of networks. Each unique environment must be evaluated against

a number of fundamental design requirements:

򐂰 Scalability to large environments: The potential growth of the network dictates the

importance of this requirement. If support is needed for large, highly redundant networks,

link state or hybrid algorithms should be considered. Distance vector algorithms do not

scale into these environments.

򐂰 Stability during outages: Distance vector algorithms may introduce network instability

during outage periods. The counting to infinity problems may cause routing loops or other

non-optimal routing paths. Link state or hybrid algorithms reduce the potential for these

problems.

򐂰 Speed of convergence: Triggered updates provide the ability to immediately initiate

convergence when a failure is detected. All three types of protocols support this feature.

One contributing factor to convergence is the time required to detect a failure. In OSPF

and EIGRP networks, a series of hello packets must be missed before convergence

begins. In RIP environments, subsequent route advertisements must be missed before

convergence in initiated. These detection times increase the time required to restore

communication.

򐂰 Metrics: Metrics provide the ability to groom appropriate routing paths through the

network. Link state algorithms consider bandwidth when calculating routes. EIGRP

improves this to include network delay in the route calculation.

򐂰 Vendor interoperability: The types of devices deployed in a network indicate the

importance of this requirement. If the network contains equipment from a number of

vendors, standard routing protocols should be used. The IETF has dictated the operating

policies for the distance vector and link state algorithms described in this document.

Implementing these algorithms avoids any interoperability problems encountered with

nonstandard protocols.

򐂰 Ease of implementation: Distance vector protocols are the simplest routing protocol to

configure and maintain. Because of this, these protocols have the largest implementation

base. Limited training is required to perform problem resolution in these environments.

In small, non-changing environments, static routes are also simple to implement. These

definitions change only when sites are added or removed from the network.

The administrator must assess the importance of each of these requirements when

determining the appropriate routing protocol for an environment.

5.1.6 For additional information

For more details on this subject, refer to:

򐂰 TCP/IP Tutorial and Technical Overview, GG24-3376

򐂰 The Basics of IP Network Design, SG24-2580

򐂰 z/OS CS: IP Configuration Guide, SC31-8775