Zeng, Xianzhi, et al. "CANDY: A Benchmark for Continuous Approximate Nearest Neighbor Search with Dynamic Data Ingestion."

Zeng, Xianzhi, et al. "CANDY: A Benchmark for Continuous Approximate Nearest Neighbor Search with Dynamic Data Ingestion."

• Category: Benchmarks & Evaluation
• Tags: benchmark, ann, dynamic-data, vector-search
• Source: arXiv:2406.19651

Description

CANDY is a benchmark introduced in 2024 for evaluating continuous approximate nearest neighbor (ANN) search systems, with a special focus on dynamic data ingestion. This is particularly relevant for assessing next-generation vector databases that must support both efficient similarity search and frequent data updates.

Features

• Provides a standardized benchmark for continuous ANN search.
• Focuses on scenarios with dynamic (frequently updated) data.
• Useful for evaluating vector database systems' performance under realistic, evolving workloads.
• Supports research and development of efficient ANN algorithms adaptable to dynamic environments.

Pricing

Not applicable; this is an academic benchmark paper.