You can configure a K2 installation to use mirroring, in which a collection is duplicated and attached to separate K2 Servers. A K2 Broker can then distribute searches and document retrievals among the mirrored collections in a load-balanced fashion.

As Figure 2-2 shows, the company in this example has decided to mirror its most heavily used collection, C1, on all three K2 Servers. (To enable load-balanced search of mirrored collections, the collection duplicates must be on different K2 Servers and must have identical names.)

This mirrored configuration allows simultaneous access to C1 by many users, distributing any performance impact. The K2 Broker handles incoming requests to search collection C1 by routing the requests among the appropriate K2 Servers.

Additionally, if one of the collections or K2 Servers goes off-line, the broker is still able to access the remaining mirrored collections, without requiring any action by the end user or administrator.

Another advantage of the distributed architecture shown in Figure 2-2 is that K2 Brokers can be duplicated for failover purposes. If the primary K2 Broker goes offline, the application can switch to the alternate and still access all the same information.


Note   You can configure K2 to automatically create mirrored collections when indexing. You can also create multiple non-mirrored collections in a single indexing pass (as illustrated by collections C2a and C2b in Figure 2-2), distributing documents between the collections based on document content or other criteria.