This section is optional. The preceding instructions are suitable for a small implementation, but will not cope well in a system with a very large number of users.
In a standard setup that we have demonstrated previously, the usage records are captured (that is, inserts) in standard tables held in the BIPlatform schema and this table is also used in the analysis (that is, reads).
This is not the most efficient way for either process, the insert or the read:
The insert record process ideally does not want to be slowed down by indexes, column constraints, and read contention
The analysis process ideally wants indexes, summaries, foreign keys, and so on, to speed up the reports
The standard out-of-the-box setup makes a compromise and puts some indexes on the table. An alternative method is to separate the two processes completely! The benefit is faster insertions of usage tracking records and faster analysis of the results. The downside is that you have to wait until the ETL...