SQL Vs NoSQL

• In general case, choose Database/SQL because they've been arround for a while
• Schema logic is both doable in Database/SQL and NoSQL
  • The only difference is for NoSQL you're moving schema to application layer rather than database layer
  • If your schema keeps changing than maybe NoSQL is better
• Choose NoSQL if your application
  • require super low latency
  • unstructured, doesn't have any relational data
  • you only need to deserialise and serialise data (JSON, XML, YAML)
  • Trash data that can be safely discarded at some points (logging, session storage)
• Choosing NoSQL is very risky. For example in MongoDB. When you write a data, the data is staged to write at a later time.
  • Now what happens if it can't write later?
  • It doesn't guarantee consistency
• Database/SQL is harder to scale horizontally because of CAP theorem but doesn't mean it's not scalable. Scalability is a very complicated topic
  • SQL follows ACID which given consistency. When horizontally scaled and network failure happens. It sacrifices Availability by terminating the whole transaction
  • Join table is harder to scale horizontally given the join table and main table are in different instances
• NoSQL is easier to scale horizontally (sharding)
  • NoSQL are contains of documents. and it's self-contained object.
  • We can put these objects on different servers without worrying about joining rows from multiple servers
  • BASE allows NoSQL to cheat the system by drop of some of the functionality to maintain scalibility concept
• Now there is NewSQL that supports Horizontally scalability for SQL

Performance

NoSQL has better performance comparing to SQL: Paper.pdf (rochester.edu)

Technical knowledge

• NoSQL uses LSM Tree (Log Structured Merge Tree)
• Database/SQL uses B Tree and B+ Tree