Endeca does not support Federated Search (at least not at this point in time). Endeca does support loading huge amounts of data into it s index and then allow for semi-realtime updates to that data.
So basically to perform the same task as a federated search you would do a few things:
- Have large imports of data from the various data sources at predetermined intervals (weekly, daily, etc.) into an Endeca Agraph.
- Have smaller, semi-realtime, data imports into the Endeca Agraph to populate any changes that might have occurred in that small timeframe. This should happen fairly often. (every 5 min, half hour, etc.).
With that setup you are able to return information from multiple systems, but those systems aren t queried real-time when a search is performed.
EDIT: This is not true as of July 2012. A main part of Endeca release at this point was the so called Endeca Assempler or Assembler API, which lets you define external systems, that are being queried at run time and returned in one request. For a programmer it is very transparent. You send a request to the Endeca engine and it will return an object containing results from all federated engines as well. It can return the data in various formats, Java, POJO, XML, JSON and some more.