When setting the support platform or even the whole warehouse, the muffler should be considered from the perspective of business. We can even investigate the existing system at the last step - if the defined indicators and dimensions are reasonable and the current R & D can not meet them, we need to supplement R & D, not modify the indicator definition, or even abandon the indicator. Data is generated from various systems, but its function is always in the business. Don't save resources. Many modelers recommend using paradigm modeling rather than dimensional modeling. This modeler is undoubtedly a technical architect - after all, there are still many systems without paradigms these days - but certainly not a data architect. From a data perspective, database capacity is always the last consideration. Or, to put it another way, tell your boss that you can increase the speed of Bi demonstration by five times and the efficiency of daily SQL writing for exhaust muffler by 10 times, but you need four times the capacity of your current hard disk. It's not even a multiple choice question for the boss. Star and snowflake. Or more accurately, it is dimensional modeling and paradigm modeling. Pure normal form modeling is not suitable for OLAP system, pure star model will encounter numerous difficulties. Based on the actual business situation (rather than the underlying system), make appropriate compromises - such as micro dimensions and support tables - to make your model truly reliable.