Milvus Project Roadmap and Time Schedule
Time | Note | |||||||
---|---|---|---|---|---|---|---|---|
2.2 | 2022.10.28 | Stable | 2.2.33.0 | 2023.28.823 | Stable | 2.3 | 2023.2.28 | Experiment|
2.4 | 2023. | 512.30 | StableUnder development | |||||
3.0 | -RC12023 | 2024. | 73.30 | Experiment | 3.0-GA | 2023.8.30 | StableUnder Design |
detailed features
Roadmap features
2.2
- Refactor Segment Assignment in Query coord
- Support Disk ANN Index
- Support RBAC
- Support BulkInsert Data
- Support Rate Limiter → Memory protection, Insertion/Search throughput
- Support collection level data TTL
- Primary/Backup coordinator
- Refine and fully tested pymilvus, java and go SDK.
- Release birdwatcher, milvus meta inspect and hacking tool
- Better Monitoring, collection level DML metrics supported
2.2.3
1. Rolling Upgrade
2. Resource group for querynode physical isolation
3. BulkInsert any sized data.
4. Performance optimization
5. Dynamic config change
2.3
- Refactor On Disk Data format
- Performance Improving on both memory/disk index
- Faster failure recovery speed.
- Merge Index coordinator/ Data coordinator into 1 role
- Integrate with new vector search engine, Knowhere 2.0
- Support Range Search
- Refine Alias functionality
- Backup support
- Dynamic Partition load/release
- Incremental data subscription
- Milvus benchmark tool
- Fully tested Restful API
- SQL support - count
- Intergated with Huggingface, OpenAI, Paddle paddle
2.4
- Support GPU index
- Support ScaNN Index
- Refactor Scalar Execution engine with Velox
- Better Hybrid search speed
- Better streaming data search performance
- Better Query pagination
- Improve multi memory replica availability
- Integrate with more inference generation system
- Tracing
- Support offline deployment
- SQL support - search,insert, delete, query
- Segment mmap load for memory index
- Index parameter optimization tool
- Compaction optimization
3.0
- Full SQL Support
- Modify collection schema dynamically - add column, remove column
- Support more datatype,map,list...
- Remove Datanode, all compaction/bulkload functionality moved into indexnode, streaming functionalities moved into log node
- Primary/Backup milvus cluster
- Support entity Update
- GPU index build
- Fully tested Cpp/Rust API
- New Milvus bootcamp
...
Category | Milvus 2.3.x (Achieved in recent releases) | Milvus 2.4.0(End of CY23 release) | Roadmap(3.0) |
Scalability and Performance | Growing Index ScaNN Index | Bulk Insert Optimazation Scalar Fields Index Holds More Collections/Partitions | Inverted Index for JSON GPU Acceleration Full test for 10 billion+ vectors |
Ease of Use | Cluster Rolling Update Delete by Expression Upsert | Support More Datatypes | Add/Delete Collection Fields Support SQL Syntax Upgrade SDK |
Enterprise Offering | RBAC Partition Key | Accesslog Enhancement Compaction Optimization | CDC for Online/Cross-Cloud Migration |
AI-augment Capability Designed features for AI applications | Dynamic Schema | Group By for Grouping Requests Multi-Vectors(beta) Sparse Vector(beta) | Scenario-oriented Vector Search Patterns Spark Connector for large scale data processing Multi-Vectors GA Sparse Vector GA |