Horizontal Scalability

In a vertically scaled database, performance depends on a single machine, leading to more CPUs, more RAM, higher cost.

In a horizontally scaled database like CrateDB, performance grows by adding more nodes to the cluster.

Each new node brings its own compute, memory, and storage resources, allowing CrateDB to:

• Ingest and process millions of records per second
• Run complex analytical queries across billions of rows in milliseconds
• Scale linearly, without performance bottlenecks

This makes CrateDB ideal for IoT data streams, AI feature pipelines, and real-time analytics platforms that never stop growing.

 

CrateDB’s scalability comes from its shared-nothing, distributed architecture:  every node operates independently, yet collaborates as part of a unified SQL cluster.

When you add a new node:

1. The cluster automatically recognizes it through node discovery.
2. Shards (the physical units of data storage) are redistributed evenly.
3. The query planner adapts instantly to include the new node in distributed queries.
4. Data replication and balancing happen in the background, with zero downtime.

Your system capacity expands immediately,  both in storage and processing power.

The diagram below illustrates the automatic redistribution process:

1. The initial three node cluster utilizes about 70% of the available storage space.
2. The addition of a new node results in an unbalanced distribution of data.
3. The automatic redistribution of data initiates, until an almost equal level of storage consumption across the four nodes is achieved again.

 

CrateDB’s distributed SQL engine scales both data ingestion and query execution linearly:

• Each node processes queries in parallel on local data.
• The coordinator node merges intermediate results.
• Adding nodes means faster response times, not slower ones.

This architecture ensures that as your data volume doubles, so does your throughput.

CrateDB scales out and back in dynamically, allowing you to adapt to workload spikes or evolving data strategies.

• Elastic scaling: Add or remove nodes without interrupting queries.
• Rolling operations: Upgrades, maintenance, and rebalancing happen live.
• Consistent performance: Automatic load balancing keeps clusters evenly distributed.

Whether you’re handling 10 GB or 100 TB, CrateDB maintains real-time query performance across the entire dataset.

ALTER CLUSTER ADD NODE '10.1.0.8';

In seconds, CrateDB redistributes shards, updates its execution plan, and starts routing queries through the new node:
no restarts, no reconfiguration, and no data movement downtime.

With CrateDB Cloud, you can even scale your infrastructure in just a few clicks:

 

Challenge	CrateDB solution
Growing data volumes	Add nodes seamlessly to increase capacity
Performance degradation under load	Linear scaling of compute and storage resources
Downtime during maintenance	Rolling rebalancing and live updates
Complex sharding logic	Automatic data distribution and replication
Unpredictable workloads	Elastic scale-out and scale-in flexibility

CrateDB’s horizontal scalability applies to all data models, not just structured tables.
The distributed SQL engine scales out uniformly for:

• Time series data from sensors and devices
• Text and document search using MATCH
• Vector similarity queries for AI and semantic search
• JSON and nested objects
• Geospatial and location-based analytics

No matter the data format, CrateDB scales it in real time.

• Start small, grow endlessly: Begin with a few nodes, scale to hundreds.
• No special configuration: Every node is equal; no primary/secondary setup.
• Predictable costs: Scale compute and storage independently, as needed.
• Future-proof: Designed for modern data growth, from IoT to AI.

CrateDB’s horizontal scalability means you’ll never outgrow your database.