Index Replication Process 

Index replication takes place in a system with master and slave index servers. The master index server manages the original indexes and the slave index servers access index copies. Replication makes sure that changes to the master indexes are transferred to the index copies.

Replication takes place in different ways depending on the type of data storage.

Replication with Decentralized Data Storage

...

The initial replication of an index takes place as follows:

...

       1.      The master index server generates an index snapshot. The name server tells the slave index servers that the index snapshot is available. The slave index servers request the snapshot from the master index server.

       2.      When all slave index servers have the index snapshot, they integrate it into their index one after the other. The slave index server currently integrating the files has the status ‘inactive’. This means that it is not available for searching. It receives the status ‘active’ again as soon as the integration has been completed.

Because all index files are copied for the initial replication, the process can take a long time if the index in question is large.

In subsequent replications the system only replicates the changed index files. This is normally a smaller amount of data than for the initial replication, and subsequent replications are therefore faster. The process flow is as follows:

...

       1.      The master index server compares the master index and the index snapshot in order to determine changed index files. It then updates the index snapshot. The name server tells the slave index servers that a new index version is available. The slave index servers request the changed index files from the master index server.

       2.      When all slave index servers have the changed index files, they integrate them into their index one after the other. The slave index server currently integrating the files has the status ‘inactive’. This means that it is not available for searching. It receives the status ‘active’ again as soon as the integration has been completed.

 

Replication with Centralized Data Storage

The initial replication of an index takes place as follows:

...

       1.      The master index server generates a complete copy of the index (index snapshot).

       2.      The slave index servers connect to the index snapshot and use this as their slave index.

If the master index changes and replication needs to take place again, the following occurs:

...

       1.      The master index server generates a second index snapshot.

       2.      The slave index servers change to the second index snapshot.

All subsequent replications take place as follows:

...

       1.      The master index server determines the changed index files by comparing the master index with the index snapshot that the slave index servers are not currently using. It then updates this index snapshot.

       2.      The slave index servers change to the updated index snapshot.