Types of Database from Blockchains to Data Lakes + Machine Learning
How to create a visual diagram that sums up all databases and put AI and blockchains in context.
Blockchains and AI via Machine Learning can be considered to sit at two ends of the spectrum of different types of databases.
- Blockchain: a one dimensional database where each entry (cell) contains a ‘checksum’ (hash), which is difficult to create, of the previous entry. Highly accurate and tamper resistant and slow to update.
- Flat file: a table with rows and columns listing multiple attributes for each entry (row). Simple.
- Relational: tables linked together so that standard attributes aren’t repeated too many times. Pragmatic database optimised for balance between updates and retrieval.
- Graph: entries (rows) that can have unlimited attributes and relationships where the attributes are different for each entry. When represented as a table, it looks like one with a potentially infinite number of columns with lots of entries blank. Good for heterogeneous things with complex relationships.
- Full text: emphasis on a precomputed index of entries so that things can be quickly retrieved (but therefore not so up to date or accurate). This type is often used for categorisation software, or search engines.
- Data lake plus machine learning: a pile of unstructured data, where features of that data can be extracted from structured filtering relationships in the retrieval. The data lake can then be batch processed to create attributes of each item in it, for putting into a regular relational database, for example.