Introducing Milvus versions 2.3.2 and 2.3.3! These latest releases bring a host of improvements aimed at enhancing the performance and user experience of our vector database system. We've carefully considered feedback from our community of developers to make these updates.
📦 PyPI: https://pypi.org/project/milvus/
📚 Docs: https://milvus.io/docs
🛠️ Release Notes:
🐳 Docker Image: docker pull
🚀 Release: milvus-2.3.3
Key highlights:
New features
🔍 Support for Array Data Types - Milvus now supports Array data types, allowing for precise metadata filtering. For example, in e-commerce, this enables advanced searches based on multiple product tags, ensuring that search results are highly relevant to user queries.
🧹 Support for Complex Delete Expressions - With Milvus 2.3.2 or 2.3.3, developers can specify detailed criteria for data removal, enabling precise cleanup, such as rolling old data or GDPR compliance-driven deletion based on user IDs. Note that deletion is not atomic; use it cautiously for precise data management.
🗂️ TiKV Integration for Metadata Storage - By integrating TiKV for metadata storage, Milvus gains improved scalability and stability. TiKV's architecture is designed to handle large-scale metadata storage efficiently, ensuring that Milvus can scale to meet the demands of growing datasets without sacrificing stability.
🌀 Support for FP16 Vector Type - Milvus's support for the FP16 vector type enhances machine learning efficiency. This data format is widely used in deep learning and ML for its ability to represent and process numerical values more efficiently, resulting in faster and more resource-efficient machine learning operations.
📊 Support for Vector Index MMAP - Go beyond mapping just the raw data 2.3.0; now you can also map the index. This feature enables you to store more data on the same machine and provides flexibility in data storage while saving costs.
Enhancement
🔄 Rolling Upgrade Experience - Milvus has strengthened its rolling upgrade feature in versions 2.3.2 & 2.3.3, providing a streamlined and efficient transition for users upgrading from version 2.2.15 in under 5 minutes.
🚀 Performance Optimization - Minimize data copy operations for optimized data loading, simplify large-capacity inserts using batch varchar reading, remove unnecessary offset checks during data padding, and address high CPU consumption in scenarios with substantial data insertions.
In addition to the highlighted enhancements, Milvus 2.3.2 and 2.3.3 offer a multitude of new features and improvements. This includes an upgraded CDC (Change Data Capture), now capable of replicating MQ (Message Queue) messages, enabling real-time data synchronization. Developers can also use the bulk insert of binlog data with partition keys, streamlining data ingestion processes. Furthermore, these versions mark the return of binary metric types such as SUBSTRUCTURE and SUPERSTRUCTURE, expanding the range of data representation options. We encourage developers to visit our release notes for a comprehensive overview of all the new features and enhancements.
Closing notes
We're steadfast in our commitment to continuously enhance Milvus, aligning with the dynamic needs of our developer community. A heartfelt thank you to our vibrant developer community whose contributions, PRs, and insightful feedback have been instrumental in shaping the roadmap of Milvus. We're eager to witness how these updates seamlessly integrate into your development projects and await your feedback. Keep coding, innovating, and shaping the future of AI! 🤖🧠
- Key highlights:
- Closing notes
Content
Start Free, Scale Easily
Try the fully-managed vector database built for your GenAI applications.
Try Zilliz Cloud for Free