NewSQL

• Relational Database that combines ACID and horizontal scaling ability of NoSQL
• Best for OLAP and OLTP
• Offer consistency > availabilty (CP from CAP theorem)
  • In consistent data can cause developer to handle edge cases much complicated than consistent environment
  • In real-world. There is no system that can guarantee 100% availability as BASE. The most availability we can guarantee is 99.99%.
    • So choosing consistency opens the door for something that's guaranteed
• How it works?
  • Using Paxos and Raft Consensus protocols to enforce consistency
    • Using a majority voting mechanism, any changes in the data required a majority of replicas to agree to the change.
    • This allows availability even though some of the nodes are down
• Some of the example
  • Citus (Postgres extension)
  • Vitess (MySQL extension)
  • Azure Cosmos DB (data can be both SQL and NoSQL)
  • VoltDB (In-memory ACID Compliant with durability techniques: Snaposhots, K-safety, replications, command logging)