Query Engine Master - Slave configuration is used to distribute the computation of replies on several servers, in order to reduce the reply time and improve scalability.
Operating Principle: One Master agent receives the query, propagates it to several Slaves agents deployed on different servers, collects the replies of those agents and integrates them to provide the answer to user query.
This configuration is an alternative to standalone search agent (afs_search), so it is also defined for a given service, status and feed triple. Reply feed is totally identical to that of a standalone Search Agent.
Note that Master - Slave configuration does not reduce disk space usage as the entire reply database has to be copied on each slave server.
MiX configuration can provide a single reply feed from different reply databases. The principle of the MiX is to mix replies and facet values from several reply databases. It produces a single 'ReplySet' with the same structure as a standard output feed from a basic single reply database.
MiX mode imposes no constraints on the reply databases that are aggregated, so it allows the aggregation of sources that are similar or different in terms of indexed documents or indexing configuration.
The Mix Mode can be installed in a distributed manner on several servers because it uses Query Engine Master - Slave architecture. In MiX configuration, Slave agents correspond to the different sources to be aggregated. Note that only one Slave agent can be configured for each source.