High Availability for NIC

You can configure high availability for NIC when you use client/server mode with the NIC host and the NIC proxies running on different machines. NIC supports several mechanisms to maintain high availability. We recommend that you use NIC replication to keep a NIC configuration highly available. NIC replication uses groups of NIC hosts that share the same configuration for NIC resolutions to respond to resolution requests.

When you use NIC in local host mode, you do not need to configure redundancy for a NIC host, because the NIC host runs within the application.

High Availability in Existing NIC Configurations

If you have a previous NIC configuration, you may be using:

• NIC host redundancy, in which a set of NIC hosts provide redundancy

  The SRC CLI does not support NIC host redundancy.

• Redundancy for SAE plug-in agents, in which a set of SAE plug-in agents provide redundancy

  If you have an SAE plug-in agent that uses agent redundancy, enable state synchronization for the agent and use NIC replication. In SRC Release 1.0.0, configuration for SAE plug-in agent redundancy is discontinued.

NIC Replication

NIC replication uses the concept of a group to identify a NIC host that has a particular configuration. A group contains one or more NIC hosts; each NIC host in a group is unique; for example, each NIC host could reside on a different system. A NIC proxy contacts specified groups that contain hosts with the same configuration to locate a managing SAE.

For example, a group might include the host DemoHost, but not two instances of DemoHost. Typically, each NIC host in a group is located in the same point of presence (POP). However, a machine can support only one NIC host. The SRC software stores groups in the directory in ou=dynamicConfiguration, ou=Configuration, o=Management, o=umc.

For example, Figure 1 shows three NIC groups with each group containing a NIC host that has the same configuration.

Figure 1: NIC Groups

Groups let you:

• Distribute network and processing load between two or more groups
• Provide failover protection if one group becomes unavailable

With NIC replication, a NIC proxy can contact multiple NIC hosts that are assigned to different groups. When a NIC proxy is configured to contact more than one group, the NIC configuration on a NIC host in each group should be equivalent—the NIC hosts should use the same configuration scenarios.

A NIC proxy selects a group by using the method specified in the configuration for the proxy; for example, the NIC proxy can randomly choose a group from a list. The NIC proxy then sends resolution requests to the corresponding host in that group. If a NIC proxy submits high numbers of resolution requests to the NIC host, you can configure the NIC proxy to randomly pick a NIC host or to pick a NIC host in a cyclic order to decrease the probability that one NIC host manages all the resolution requests.

Figure 2 shows resolution requests sent by means of a round-robin selection.

Figure 2: NIC Group Selection by Round-Robin

If the NIC host fails to respond to a specified number of resolution requests, the NIC proxy stops sending resolution requests to the unavailable NIC host and sends the resolution requests to another NIC host. The NIC proxy continues to poll the unavailable NIC host to determine its availability. When the NIC host becomes available, the NIC proxy can again send resolution requests to that host.

Figure 3 shows a NIC proxy that sends a resolution request to Group 1, receives an error message, then sends two more resolution requests before sending a request to Group 2 rather than Group 1. When Group 1 is available again, the NIC proxy will send the request to Group 1.

Figure 3: NIC Resolution Request

You configure NIC replication for hosts, then configure NIC proxies to use replication.

Although you can distribute agents and resolvers among different hosts, as shown in the configuration for the NIC hosts OnePopBO and OnePopH1 in the sample data, we recommend that you use the DemoHost configuration, which centralizes the configuration for agents and resolvers.