Clustering Essbase databases enables load balancing and failover support. Provider Services provides parallel clustering, in which a series of active, duplicate databases respond to user requests. Which database is accessed is transparent to users, who connect to and retrieve data from one data source. Provider Services facilitates the routing of connections between databases in a cluster, based on availability and precedence rules.
In Figure 3, Essbase Database Clustering with Provider Services, Smart View users connect to Essbase through Provider Services.
Each user connection is assigned to a server during the Essbase session. Provider Services uses session-level load balancing. For example, in Figure 3, Essbase Database Clustering with Provider Services, User 1’s connection is mapped to Data Source A. User 2’s connection is mapped to Data Source B. User 3’s connection is mapped to data source C. All requests from User 1 are handled by Data Source A for the duration of the connection.
If data source A fails:
User 1 times out at Data Source A.
User 1 is rerouted to the next available data source, which is Data Source C in Figure 4, Database Cluster with One Data Source Offline.
Figure 4, Database Cluster with One Data Source Offline illustrates what happens when Data Source A goes offline.
In Figure 4, Database Cluster with One Data Source Offline, the state of query 1 is maintained at the middle tier and rerouted. Provider Services also provides load balancing across servers.
Figure 5, Essbase Database Cluster on One Server depicts clustered databases deployed on one server.
In Figure 5, Essbase Database Cluster on One Server, two servers contain Essbase databases. Server 1 has four processors and 8 GB of RAM. Server 2 has eight processors and 16 GB of RAM. Because Server 2 has more resources, it contains Data Sources B and C. Therefore, Server 2 can handle both connections.
Failover support also applies for database clusters on one server. In Figure 6, Failover for Database Cluster on One Server, Server 2 goes offline. User 2 and User 3 are then rerouted to the next available server, Server 1.
