#PostGresSQL conf 2016: #SQL vs #noSQL
Posted on April 18th, 2016
04/18/2016 @ New York Marriott Brooklyn Bridge
The afternoon panel was composed of vendors providing both SQL and noSQL database access. The discussion emphasized that the use of a #SQL vs #noSQL database is primarily driven by
- The level of comfort developers/managers have in a SQL or noSQL
- Whether the discipline in SQL rows and fields were useful in creating applications
- Whether applications use JSON structures which are easily saved in a noSQL database
- Linking to existing applications can be done either using a SQL database or using an ORM to a noSQL database.
The available of ORM (Object-relational mapping) software blurs the lines between SQL and noSQL databases. However, one is advised to avoid using an ORM when using a noSQL database initially so one can gain familiarity in the differences between SQL and noSQL.
Both db types need planning to avoid problems and depends on the situation. For instance, sharding might be best done late when there is only a single application being developed. However, if one has many applications using the same infrastructure, one should consider specifying sharding policy early.
People want to avoid complexity, but don’t want to delegate setup to a standard default or procedure.
Controlling access can be done (even if there is no data in the object) by creating views which are accessible only by some users.
Geolocation data is best handled by specialized db’s like CardoDB: The coordinate system is not rectangular and data can be handled by sampling and aggregation.