NewSQL
NewSQL is a class of database systems that aim to provide the horizontal scalability of NoSQL databases while retaining the ACID transactional guarantees and familiar SQL interface of traditional relational databases.
Definition
NewSQL is a class of database systems that aim to provide the horizontal scalability of NoSQL databases while retaining the ACID transactional guarantees and familiar SQL interface of traditional relational databases.
Overview
The rise of NoSQL databases in the 2000s offered a solution to scaling problems that traditional relational databases struggled with, but it typically required giving up strong consistency, SQL, or full transactional guarantees, as described by the CAP theorem. NewSQL systems emerged to close this gap, using new distributed architectures — often based on consensus protocols like Raft or Paxos and automatic data sharding — to scale out horizontally across many nodes while still guaranteeing ACID properties and supporting standard SQL. Technically, NewSQL databases typically achieve this through transparent sharding (splitting tables across nodes automatically, without requiring the application to manage partitioning logic), distributed consensus for coordinating transactions and replication consistently across nodes, and a SQL query layer that hides this distributed complexity from application developers, who write queries much as they would against a traditional single-node relational database. This makes NewSQL attractive for applications that need both the scale of a distributed system and the strong consistency and relational modeling that many NoSQL databases sacrifice — financial systems, global user-facing applications, and any workload where eventual consistency is not acceptable but a single-node database can no longer handle the load. Notable NewSQL systems include Google Cloud Spanner (globally distributed with strong consistency via synchronized clocks), CockroachDB (inspired by Spanner, open source), TiDB, and YugabyteDB, each offering variations on distributed SQL with strong consistency guarantees.
Key Concepts
- Combines horizontal scalability of distributed systems with full ACID transactions
- Supports standard SQL, unlike most NoSQL databases
- Uses transparent sharding so applications don't manage partitioning manually
- Relies on distributed consensus protocols (e.g. Raft, Paxos) for consistency across nodes
- Designed to avoid the consistency tradeoffs many NoSQL systems accept under the CAP theorem
- Well suited to globally distributed applications needing strong consistency at scale