The high availability and performance of Map Intelligence is addressed by the quality of the network, access permissions, and the power of hosting machinery and related software such as the database, BI tool and mapping server being used. It's impossible to generalise because the network load is dependent on the complexity of the individual requests being made by the end-user and the design of the underlying database - BI architecture, ie: the quantity and timing of data movements.
Map Intelligece requires a Java Virtual Machine and is independent of the operating system. A dual core 3.2GHz processor with 1Gbyte of RAM set aside for Map Intelligence will support 20 concurrent users with good response times. The system should be scaled appropriately for more users, or multiple computers deployed with load sharing. Because most of the work is done in memory, the amount of memory needed is highly dependent on the complexity of the requests. Map Intelligence scales reasonably linearly with CPUs and it is usually memory that is the limitation, so it is better for larger installations to use load sharing with multiple machines. Allow about 100Mbytes disk space for the Map Intelligence server installation (without the map provider) and 200Mbytes for a Map Intelligence for GeoServer installations. Map data space is additional to this. A typical Postgresql-postgis installation with the full PSMA (Australian) data set takes about 11Gbytes.
All the "standard" user interface methods are used by Map Intelligence such as rotating icons indicating waiting for data or processing to complete. Time-outs are informed via pop-up messages. Connection failure is typically handled outside of Map Intelligence.
Map Intelligence is regularly run in Virtual Machines.