heavyai/heavydb

What it solves

HeavyDB is a SQL-based, relational, columnar database engine designed to handle multi-billion row datasets. It eliminates the need for indexing, pre-aggregation, or downsampling, allowing users to query massive datasets in milliseconds.

How it works

The engine leverages the full parallelism of modern hardware, utilizing both CPUs and GPUs (specifically Nvidia GPUs) to accelerate queries. It employs a multi-tiered caching system that moves data between storage, CPU memory, and GPU memory, and uses a Just-In-Time (JIT) query compilation framework to maximize performance.

Who it’s for

It is built for users and organizations dealing with massive-scale data analysis where millisecond response times are required without the overhead of traditional database indexing.

Highlights

• SQL-based relational columnar database.
• GPU-accelerated querying for extreme performance.
• Supports multi-billion row datasets without indexing or downsampling.
• Multi-tiered caching across storage, CPU, and GPU memory.
• JIT query compilation framework.