Pregunta de entrevista de BST Global

What would your database structure look like if _____?

Respuesta de la entrevista

Anónimo

10 sep 2022

Hi! I love working on this systems. Also have been able to work with machine learning and deep IA in the ERP field. Currently i'm employed in Spain. Well now to the funny point! What would your database structure look like if you would like to have an on-premise or cloud based ERP system? -It would be a relational database like Oracle or SQL Server or PostgreeSQL or MySQL. Every ERP needs to have structured data that has some link with other data (for example invoices with taxes, orders, returns,customers,vendors,accounts, etc.). A relational database is the fastest database structure to achieve this. Snowflake one could be a good choice but ERP systems normally needs a many-to-many relation like in the example given so relational DB outperforms it and other reation-based schemas like the Star one. Whether it's an on-prem or cloud system choice, the best ERP database scheme is still a relational database because of the data type that an ERP system manages. Still if the company ERP manages some unstructured data, like audio or video, it could be wise to have the main system with a relational database that have some table with the necessary data that makes possible to retrieve the unstructered data stored on other type of system. For unstructered datal like the best would be a NoSQL with key-value scheme BUT it's not recommended to use different DB together on a system so I would encourage storing such kind of data directly on a fast file system, making a decoding method that would accelarete the process of retrieval based on the relational database table key that's querying the audio/video file. For other type of data like real-time data the system must be really isolated so the best option would be to have two different systems, one with the ERP having a relational database and other storing the real-time data of the sensors having them realte by common pre-known variables like time and/or name of the sensor and/or location, etc. Then with some API make it possible to communicate the raw data (API based on sensor NoSQL system) or extract, transform and load it to the ERP system (ET on the sensor side and L using API that let you access relational database directly)