whichdb

HEAVY.AI (HeavyDB)

active

by NVIDIA (HEAVY.AI) · since 2017 · also known as HeavyDB, OmniSciDB, MapD, HEAVY.AI

GPU-native columnar SQL database with visual analytics, fully open-sourced by NVIDIA in 2026.

columnar / olaprelationalspatialdatabase engineopen source (permissive)mature
Website Docs Source★ 3.1K Compare Find similar

Overview

HeavyDB (formerly MapD and OmniSciDB) is an in-memory columnar SQL engine that compiles queries with LLVM and runs them across GPUs, making interactive scans and geospatial joins over billions of rows feasible on a single server. NVIDIA acquired HEAVY.AI in 2025, and in October 2026 released v10.0 with the whole platform open source under Apache 2.0, including the Immerse dashboard UI, the GPU rendering engine, HeavyConnect foreign tables and HeavyRF. It is a strong option for geospatial and telemetry visualization on NVIDIA hardware; it needs GPUs and the community is still forming around the new release.

Analytics (OLAP)GeospatialReal-time analytics

Strengths & trade-offs

Strengths

  • +GPU execution makes billion-row scans and geo joins interactive
  • +Server-side GPU rendering of points, polygons and heatmaps
  • +Now fully open source including the Immerse visualization UI
  • +Native geospatial types and functions
  • +Backed by NVIDIA hardware and CUDA expertise

Trade-offs

  • −Requires NVIDIA GPUs for its main advantage
  • −Working set must fit in GPU/CPU memory
  • −Transactions and updates are limited
  • −Community and roadmap are in flux after the acquisition

Great for

  • +Geospatial visual analytics
  • +Telecom and RF network planning
  • +Interactive exploration of large telemetry datasets

Think twice for

  • −OLTP
  • −CPU-only environments
  • −Petabyte-scale cold data

Capabilities

Native or official first-party support · 7 of 25 tracked

✓Joins
✓Geospatial
✓Columnar storage
✓In-memory
✓Durable persistence
✓Compression
✓Role-based access
–Transactions
–Secondary indexes
–Full-text search
–Vector search
–JSON
–Change data capture
–Triggers
–Stored procedures
–Materialized views
–TTL / expiry
–Pub/Sub
–Encryption at rest
–Time travel
–Branching
–Online schema changes
–Streaming ingest
–Offline sync
–Runs in WebAssembly

Scale & consistency

Comfortable data size
Medium · ≤ 10 TB
Horizontal scaling
Native
Multi-region
No
Transactions
Limited ACID
Consistency
Strong
Schema
Strict
Architecture
Single node, Distributed, In memory
Query languages
SQL

Best on one or a few multi-GPU servers; hot data must fit in GPU/CPU memory.

Deployment

Self-hosted Vendor-managed cloud Serverless Embedded Edge Kubernetes operator

Ecosystem

apis & protocolsThrift API, JDBC, ODBC, Arrow
client languagesPython, Java, JavaScript, C++

Alternatives to HEAVY.AI (HeavyDB)

Compare all→
KineticaColumnar / OLAP6.4

GPU- and vector-accelerated database for real-time spatial, time-series and vector analytics.

heavydb vs kinetica →
ClickHouseColumnar / OLAP69.8

Open-source columnar OLAP database known for very fast aggregations over billions of rows.

heavydb vs clickhouse →
DuckDBColumnar / OLAP63.7

In-process columnar SQL database: the SQLite of analytics, running inside your app or notebook.

heavydb vs duckdb →
Apache DruidColumnar / OLAP35.0

Real-time analytics database for sub-second slice-and-dice over streaming event data.

heavydb vs apache-druid →
OcientColumnar / OLAP3.8

Hyperscale data warehouse on NVMe compute-adjacent storage, rebranded as the OcientAIQ platform.

heavydb vs ocient →
InMemory.NetColumnar / OLAP0.2

Commercial in-memory columnar SQL engine for .NET that speeds up reporting over hundreds of millions of rows.

heavydb vs inmemory-net →

HEAVY.AI (HeavyDB) FAQ

What is HEAVY.AI (HeavyDB)?

HEAVY.AI (HeavyDB) is a columnar / OLAP database by NVIDIA (HEAVY.AI), first released in 2017. HeavyDB (formerly MapD and OmniSciDB) is an in-memory columnar SQL engine that compiles queries with LLVM and runs them across GPUs, making interactive scans and geospatial joins over billions of rows feasible on a single server. NVIDIA acquired HEAVY.AI in 2025, and in October 2026 released v10.0 with the whole platform open source under Apache 2.0, including the Immerse dashboard UI, the GPU rendering engine, HeavyConnect foreign tables and HeavyRF. It is a strong option for geospatial and telemetry visualization on NVIDIA hardware; it needs GPUs and the community is still forming around the new release.

Is HEAVY.AI (HeavyDB) open source?

Yes. HEAVY.AI (HeavyDB) is open source. License: Apache License 2.0.

Does HEAVY.AI (HeavyDB) support ACID transactions?

Partly. HEAVY.AI (HeavyDB) offers limited ACID guarantees (for example, single-record or single-partition transactions), with strong consistency.

Does HEAVY.AI (HeavyDB) scale horizontally?

HEAVY.AI (HeavyDB) scales horizontally out of the box, with built-in sharding or partitioning. It is comfortable at medium scale (≤ 10 TB).

How can I run HEAVY.AI (HeavyDB)?

You can self-host it.

What is HEAVY.AI (HeavyDB) best for?

HEAVY.AI (HeavyDB) is a strong fit for geospatial visual analytics; telecom and RF network planning; and interactive exploration of large telemetry datasets. Think twice for OLTP.

What are the best alternatives to HEAVY.AI (HeavyDB)?

Common alternatives to HEAVY.AI (HeavyDB) include Kinetica, ClickHouse, DuckDB, Apache Druid and Ocient. The right choice depends on your workload, scale and hosting; WhichDB's smart match ranks them against your requirements.

How popular is HEAVY.AI (HeavyDB)?

HEAVY.AI (HeavyDB) ranks #234 of 844 databases on the WhichDB Score (14.3/100), and #24 among columnar / OLAP databases. The score combines public adoption signals such as package downloads, GitHub activity, job posts and developer discussion.