Scale Agentstore and Utility IDOLs

The previous sections have discussed the sizing and scaling of a distributed IDOL index. However, you can also deploy the IDOL Content component in situations that do not have a large associated index, for example as an Agentstore (when the number of documents, and the documents themselves, tend to be smaller), or for performing utility functions, such as language detection, document statistics, or highlighting.

In these circumstances, it is not generally necessary to distribute documents, but you might need to distribute queries using a single DAH in mirror mode. You can use a single DIH, also in mirror mode to keep the contents of the child servers (for example agents) in sync. In this case, you can scale the system by assessing the throughput of a single Agentstore, again most practically by realistic small-scale testing.

If multiple parallel processes might use the child servers (for example, multiple instances of CFS run on several machines), it might make sense not to route all queries through a top-level DAH. In this case, you might instead want to connect each CFS or other component to a local Agentstore or utility IDOL running on the same machine.