Using dynamic or static routing to improve network selection

You can use the IPv6 OSPF or IPv6 RIP dynamic routing protocol provided by the OMPROUTE routing daemon to provide information about the IPv6 prefixes and hosts that can be accessed indirectly by way of adjacent routers. You can use IPv6 OSPF or IPv6 RIP, either alone or together with IPv6 router discovery, to provide complete routing information.

For routing considerations for interfaces that use the OSX CHPID type, see the information about OMPROUTE considerations for the intraensemble data network in z/OS Communications Server: IP Configuration Guide.

When both of the following statements are true, only default routes are available for accessing hosts that are not on directly attached links:

Neither the IPv6 OSPF dynamic routing protocol nor the IPv6 RIP dynamic routing protocol of OMPROUTE is being used.
Adjacent routers are not including indirect prefix routes (using the Route Information option as described in RFC 4191 Default Router Preferences and More-Specific Routes) in their router advertisement messages.

If the TCP/IP stack uses a non-optimal router when data is sent to one of these hosts, that router can send a redirect message that indicates a more optimal router for future use, as long as the more optimal router is on the same LAN as the original router.

When the TCP/IP stack is connected to multiple LANs, this processing might result in the following situations:

A non-optimal router is used
A router is used that cannot reach the final destination

For example, if the stack selects a router on one LAN, but the optimal router is on another LAN, the router on the first LAN cannot redirect the stack to the router that is on the second LAN. In this case, configure a static route so that the stack can initially select the optimal network path.

Guidelines: When you are defining static routes, use the following guidelines:

Use subnet routes instead of host routes
Remote IP addresses are difficult to predict. When using extensions to stateless address autoconfiguration, some clients can change their IP addresses on a routine basis, such as once an hour or once a day. In addition, these addresses can be created using cryptographic algorithms, making it difficult to impossible to predict which IP address a client might use. Defining static host routes to be used when communicating with such a client is equally as difficult or impossible.

Instead of defining a host route, define subnet routes. The network prefixes used in generating IPv6 addresses are much more stable than the interface identifiers used by hosts, typically changing only when a site is renumbered.
Use the link-local address of gateway router
When you are defining the gateway router for a static route, use the link-local address for the router. Link-local addresses do not change as the result of site renumbering, which minimizes potential updates to the static routes. This is required in order to honor and process an ICMPv6 redirect message.
Effects of site renumbering on static routes
When a remote site is renumbered, new network prefixes are defined for the remote site and the old network prefixes are deprecated. After a time period, the old network prefixes are deleted.

A static route to a remote subnet should be created when a prefix is defined and should remain as long as the prefix is either preferred or deprecated. Only when the remote prefix is deleted should the static route be deleted.