Notes about databases.
A common standard interface, published in 1969, for access to databases supporting network data model using tuple-at-a-time queries. These databases give developers precise low-level control of data representation. One disadvantage is that the format of data storage is a part of public API and is impossible to change without rewriting client code. Another disadvantage is that network data model causes complex queries such as manually joining data using nested for loops.
Published in 1970 by Edgar Frank "Ted" Codd, designed to abstract away physical data representation and instead present a high-level relational interface to access data as a set of tables.
An attempt to avoid object-relational impedance mismatch, they store objects directly. Main downsides are a lack of a single expressive query language and inability to use from different programming languages.
Databases specifically build to run OLAP (analytical) workloads efficiently. Don't support OLTP (transactional) workloads well, which is worked around by transferring slightly stale data from the main transactional database and using the analytical database purely for analytics. The main way to gain performance is by using columnar storage.
A loose movement of databases deciding to forgo the relational model, SQL and ACID principles to more easily achieve high performance and scalability by sharding. Another common characteristic is lack of schemas.
A distinct family of databases, that all tend to be non-relational, designed specifically to ingest and store a massive number of data points with a time component. The end purpose is to answer analytical queries on these temporally ordered data points.
Because of the typical amount and nature of data (billions of points per day), compression becomes very important. On the other hand, updates to already ingested data and transactional workloads are rare and so their performance is not prioritized.
A new generation of relational databases supporting SQL that attempt to achieve scalability and performance on transactional workloads that are comparable with those of NoSQL databases.
A big trend among modern databases is ability to support hybrid (HTAP) workloads.