In multi-homed situations, RPC Manager will facilitate communications where multiple network interfaces appear on a workstation, provided each network interface has its own IP address.
You can configure the RPC Manager to use the available networks in two ways:
• Prioritized; or
• Round-robin.
These two methods can be described as follows:
Prioritized
The available networks are arranged in an order of priority. The RPC Manager always uses the highest priority available network; or
Round-robin
All network interfaces are treated equally, and RPC Manager traffic is sent using each interface in turn. This method could be used to improve the tolerance of the distributed system for network failures, and is the default behaviour.
The above two methods can be combined where there are three or more networks, such that one of the networks could be prioritized and the other two left as round-robin. RPC Manager traffic will always be sent over the prioritized network first, using the remaining two networks in round-robin mode if the prioritized network link to a workstation fails.
To establish the prioritized mode, you can enter the IP mask for each network, (in order of priority) followed by a number of bits representing the number of sections of the IP to use to determine the subnet.
For example:
[RPCManager-NetPriority]
IP = 192.168.1.0/24
IP = 192.168.2.0/24
In the above example, the variables specify two IP masks using the form:
<IP address>/<number of bits>
such that the number of bits specified is applied to the IP address from the most significant end. A mask value of 24 specifies the first three numeric parts of the IP address (8-bits per number), a mask of 16 the first two numeric parts, and so forth.
The example above specifies that IPs on subnet 192.168.1 are to be given priority over IPs to the same workstation on all other subnets. Likewise, IPs on subnet 192.168.2 are to be given priority over IPs to the same workstation on all other subnets, except subnet 192.168.1. If an IP on subnet 192.168.1 becomes unusable, the corresponding one on subnet 192.168.2 will be used. If that also fails, any remaining IPs on other subnets will have RPCs transmitted to them in round-robin fashion.
For further information on these methods and multi-homed systems and RPC, see Multi-homed Systems.