DLSw+ Redundancy and Fault Tolerance
Problem
You want to improve the fault tolerance of your DLSw network.
Solution
There are several things you can do to improve the reliability and fault tolerance of your network. Many of these solutions have the added benefit of improving performance. The first important thing to consider is having more than one DLSw peer router connected to the mainframe's Token Ring. In this case, you will want to make sure that you balance the load between the two peers as much as possible:
dlsw-branch#configure terminal
Enter configuration commands, one per line. End with CNTL/Z.
dlsw-branch(config)#source-bridge ring-group 101
dlsw-branch(config)#dlsw local-peer peer-id 10.1.2.5
dlsw-branch(config)#dlsw timer explorer-wait-time 5
dlsw-branch(config)#dlsw load-balance circuit-count
dlsw-branch(config)#dlsw remote-peer 0 tcp 10.1.1.5 lsap-output-list 200
dlsw-branch(config)#dlsw remote-peer 0 tcp 10.1.1.9 lsap-output-list 200
dlsw-branch(config)#dlsw allroute-sna
dlsw-branch(config)#end
dlsw-branch#
Discussion
This example is extremely similar to the one shown in Recipe 15.2, so the interface parts of the configuration are left out for simplicity. The main difference is the presence of a second dlsw remote-peer command. This command points to the IP address of a second router that is on the same central Token Ring as the mainframe. So the same mainframe MAC address is visible on this branch router coming from both DLSw peers.
The command dlsw load-balance circuit-count tells the router to balance circuits between these two peers. For example, if several PCs are connected the branch Token Ring, and they all have SNA sessions to the mainframe, this feature will ensure that half of these sessions follow one path and half follow the other. If one of these peers fails, then the circuits will be re-established through the remaining path.
The routers will not tear down circuits to balance the load, but rather will look at the current number of circuits on each peer each time a new circuit is established. They will then use this information to decide where to put the new circuit.
The other option here is the default round-robin circuit balancing. Round robin has the disadvantage that it takes a very long time to rebalance the load after a failure. So we recommend using the circuit-count option for load balancing between two or more remote DLSw peers.