Distributed SQL (NewSQL)

What Is NewSQL?

NewSQL databases combine the scalability of NoSQL with the ACID guarantees and SQL interface of traditional relational databases. They distribute data across multiple nodes while maintaining strong consistency and full SQL support. This eliminates the choice between scale and consistency.

How Distributed SQL Works

Distributed SQL databases use consensus protocols (Raft, Paxos) to replicate data across nodes. Each table is divided into ranges, and each range is replicated across multiple nodes. Transactions span multiple ranges using two-phase commit or parallel commits.

SQL Query: UPDATE accounts SET balance = balance - 100 WHERE id = 1

┌─────────────┐
│   SQL Layer  │  Parses, plans, optimizes
└──────┬──────┘
       │
┌──────▼──────┐
│ Transaction  │  Coordinates distributed transaction
│    Layer     │  Two-phase commit across ranges
└──────┬──────┘
       │
┌──────▼──────┐     ┌─────────┐     ┌─────────┐
│  Range 1    │────►│ Range 2 │────►│ Range 3 │
│ (Raft Group)│     │(Raft Grp)│     │(Raft Grp)│
│  3 replicas │     │ 3 replicas│    │ 3 replicas│
└─────────────┘     └─────────┘     └─────────┘

CockroachDB

CockroachDB is a distributed SQL database designed for global applications. It automatically shards data, handles rebalancing, and provides serializable isolation. It is wire-compatible with PostgreSQL, making migration straightforward.

• Automatic range splitting and rebalancing as data grows.
• Geo-partitioning for data locality and compliance.
• Serializable isolation by default.
• PostgreSQL wire protocol — use any Postgres driver or ORM.

Google Spanner

Spanner is Google's globally distributed database with external consistency (stronger than linearizability). It uses TrueTime (atomic clocks + GPS) to order transactions globally without coordination. Spanner is the gold standard for distributed SQL but is only available on GCP.

TiDB

TiDB is an open-source distributed SQL database compatible with MySQL. It separates compute (TiDB server) from storage (TiKV), allowing independent scaling. TiDB is popular in the Chinese tech ecosystem and handles HTAP (hybrid transactional/analytical) workloads well.

Tradeoffs

• Higher latency per query — Distributed transactions add 10-50ms overhead.
• Operational complexity — More nodes to manage, monitor, and upgrade.
• Cost — More infrastructure than a single database instance.
• Limited ecosystem — Some SQL features (certain joins, stored procedures) may not be supported.