# FreshDiskANN

> Fast and accurate graph-based ANN index for streaming similarity search, enabling real-time updates on billion-point indexes using a single machine with real-time freshness.

- **URL:** https://arxiv.org/pdf/2105.09613
- **Category:** Research Papers & Surveys
- **Tags:** Ann, graph-based, dynamic-updates
- **Updated:** 2026-03-10 11:58
- **Canonical page:** https://vectordb.ever.works/items/freshdiskann

## Details

## Overview

FreshDiskANN is a fast and accurate graph-based ANN (Approximate Nearest Neighbor) index designed for streaming similarity search, aiming to serve and update an index over a billion points with real-time freshness using just one machine.

## Key Features

- Billion-scale index support on single machine
- Real-time index updates
- Graph-based indexing approach
- Disk-based storage for large datasets
- Streaming data ingestion
- High accuracy with fast queries

## Technical Approach

- Graph-based nearest neighbor search
- Optimized for disk-resident data
- Efficient update mechanisms
- Balances accuracy and speed
- Single-machine scalability

## Use Cases

- Real-time recommendation systems
- Streaming data applications
- Large-scale similarity search
- Dynamic vector collections
- Continuous data ingestion scenarios

## Performance

- Handles billion-point datasets
- Real-time freshness guarantees
- Fast query response times
- Efficient update operations
- Single-machine deployment

## Research Context

Part of the DiskANN family of algorithms from Microsoft Research, focused on making large-scale vector search practical on commodity hardware with disk-based storage.

---

_This Markdown mirror is generated by Ever Works for AI agents. The canonical HTML page is at https://vectordb.ever.works/items/freshdiskann._