The big advantage of static routing is the simplicity. A static route identifies a destination and the appropriate link to take to reach that destination. Static routing usually takes advantage of default routes: when the destination is not explicitly coded, send the packet to the default router and let that router figure out how to get the packet to its destination.

So static routing is easy--but not very resilient. Most installations would not use static routing. Instead, a dynamic routing protocol such as OSPF would be used. Dynamic routing takes a little more effort to plan and set up, but once the planning stage is completed, the network effectively takes care of itself. It can also make more efficient use of network topology: instead of dumping everything to a default router, OSPF can take advantage of more intelligent network design.

For testing purposes and for smaller installations, static routing might be all that is required. There are also situations where an organization would use a combination of static and dynamic routes.

In z/OS, the statement block used to configure static routes is the BEGINROUTES statement. In Figure 1, a sample of static routes that could be used for LPAR 1 are coded.

 BEGINROUTES
 ROUTE 201.2.11.0 255.255.255.0 =            OSAELNK1 MTU 1500
 ROUTE 201.2.11.0 255.255.255.0 =            OSAELNK2 MTU 1500
 ROUTE DEFAULT                  201.2.11.100 OSAELNK1 MTU
DEFAULTSIZE
 ROUTE DEFAULT                  201.2.11.100 OSAELNK2 MTU
DEFAULTSIZE
 ENDROUTES

Figure 1 begins with two equivalent routes for reaching the directly-attached subnetwork 201.2.11.0. Following that are two default routes pointing to a router (not shown) with IP address 201.2.11.100.

Some redundancy is built into this scenario: either adapter can be used to reach the directly attached network and either adapter can be used to reach the default router. It sounds like static routes have some resiliency, right?

Well, the difficulty is in the fact that the second adapter is only used if TCP/IP on z/OS detects a failure with the link. For example, if OSAELNK1 were to be stopped manually, TCP/IP would immediately switch to using OSAELNK2.

However, if a problem occurred with one of the adapters and the problem was not detected by the OSA-Express card or by TCP/IP itself, then the adapter would continue to be used. It's better than nothing, of course, but a routing protocol like OSPF would automatically sense and make accommodations for the bad link.

The other difficulty with static routes occurs with VIPA when it is in a different subnetwork from the physical interface. How does an adjacent router (for example, 201.2.11.100 as shown in Figure 1) know that 198.2.11.9 in LPAR 2 is to be reached by sending a packet to 201.2.11.3 or 201.2.11.4? The answer is that the router needs to be told by coding a static route in its configuration. Again, in this context "OSPF is your friend."