If you are not dedicated to cloud storage, but to cloud applications, generally, application systems with slightly more complex business relationships require the combination of database and unstructured (nosql) methods, log Data and a large amount of file data are managed in an unstructured manner and used as specialized file servers. structured data needs to be managed in a database, but structured data needs to be classified, for example, it can be divided into Dictionary data, basic data, and business data. For Dictionary data and basic data, a common method can be used for a table. For business data, the operation time segment is particularly strong, generally, operations are less likely to be performed over time. Business Data over two years will basically become silent data, but will only be queried occasionally or used for data mining. Therefore, business data can be stored in time-based table sharding or database sharding to ensure the processing performance of the current data. This method is actually very old. This method was used when we first started ERP. Each business form is divided into the current table and History Table (which can also be stored on a yearly basis ), as long as the data has been closed, it has been moved from the current table to the History Table for more than a year (with an automatic tool, which is automatically completed every year), thus improving the performance. Of course, you can customize the actual business system based on the business and data features of your system.
Of course, cloud applications generally adopt multi-tenancy and can be stored by enterprise partition (because data between enterprises is independent ), this reduces the data storage and throughput pressure on a node.