IDEA

Category: SDKs & Libraries
Website/Source: https://github.com/asaflevi0812/IDEA

Description

IDEA (Inverted Deduplication-Aware Index) is a research implementation of an index structure designed to improve storage efficiency and query performance for similarity search workloads. It targets high-dimensional vector data and content-addressable data, making it applicable to large-scale vector databases and approximate nearest neighbor (ANN) indexing systems. The codebase accompanies the paper “Physical vs. Logical Indexing with IDEA: Inverted Deduplication-Aware Index” (FAST ’24) and builds on the Destor open-source storage system.

Features

• Inverted, deduplication-aware index structure for similarity search workloads.
• Support for high-dimensional vector data and content-addressable data.
• Research implementation of naïve and deduplication-aware indexes, based on the Destor storage system.
• Designed for large-scale vector database / ANN indexing scenarios, focusing on storage efficiency and query performance.
• Configurable via repository-provided configs, dataset details, and keyword definitions.
• Build and automation scripts for compiling the code and generating build commands.
• Dependency installation script (install_dependencies.sh) tailored to a clean Ubuntu Server 22.04 LTS environment.

Technical Requirements

• Operating system: Ubuntu Server 22.04 LTS (clean image recommended).
• Dependencies: Installed via the provided install_dependencies.sh script.
• Version control: Git required to clone the repository.

Pricing

• Open-source research code (no pricing information provided).