In this scenario, two or more identical pieces of hardware serve as Crowd nodes in the cluster. These servers run the Crowd application and host the Crowd installation and home directories; each node must be configured identically. HAProxy, a session-aware load balancer, is used to direct traffic between the available nodes.
The RDBMS uses native mirroring (replication) technology to provide a spare (slave) database server. Note that the choice of RDBMS will largely dictate the implementation of the mirroring and monitoring, with each RDBMS having greater or lesser capabilities. For example, MySQL provides mirroring but not monitoring, so an off-the-shelf or custom monitoring solution is necessary.