Planning System Model

The system model is the arrangement of virtual units on physical servers. Virtual units are logical units with a virtual IP address. Placing services in different virtual units and using the High Availability Control (HAC) service ensures that one virtual unit continues running if another one is stopped for example for updating, or by a server failure.

Virtual units should have instances on at least two different physical servers. During a possible server failure the services are automatically moved to other servers by HAC. If the system should withstand a failure of more than one server at a time, more instances of each virtual unit should be created on different servers. Spreading the primary instances on different servers enables load balancing in normal conditions when no virtual unit instance has failed. Load balancing should also be taken into account when arranging the backup instances.

Each customer must have their own virtual units, but the virtual unit instances of different customers can be placed on the same physical server. When deciding which customers could share a server, consider the following:

• Similar customers can be placed on the same servers. For example, customers not making frequent upgrades can be placed on same servers together. But less stabile customers, such as the ones that always want the latest upgrade or frequently have new customizations, should be on different servers. This decreases the potential downtime due software upgrades.

• Customers with a similar maintenance window can easily share a server, but customers with non-overlapping maintenance windows should be placed on different servers.