This page provides an overview of new benefits and feature upgrades we hope to provide for users in our upcoming releases, as well as our long-term goals. It is driven by user priorities and may change as their requirement or demand changes.
Milvus roadmap.
Milvus 2.0 time schedule
Upcoming major releases:
Milvus 2.1: November 2021
Milvus 2.2: February 2022
Roadmap features
DDL
Version
Feature
Owner
Status
Comment
2.0.0-RC
Supports numerical scalar data types
done
2.0
Supports string data types
czs007, dragondriver
in progress
2.0
Collection alias
lsgrep
in progress
2.1
Supports Scalar bitmap/inverted Index for string and numeric data types
pending
2.1
Supports data life cycle management
pending
2.1
Automatic data partition
pending
2.2
Collection rename
pending
DML
Version
Feature
Owner
Status
Comment
2.0.RC
Supports scalar filtering
done
2.0.0-RC
Supports for query by primary key
done
2.0
Supports query by expression
fishpenguin
in progress
2.0
Supports delete by primary key
scsven
in progress
2.1
Supports search by primary key
pending
2.1
Vector similarity search by distance
pending
2.2
Supports search/query result pagination
pending
2.2
Supports upsert/primary key deduplication
pending
Features
Version
Feature
Owner
Status
Comment
2.0.0-RC
Supports time travel to any specified point in time
done
2.0.0-RC
Offers three levels of tunable consistency: strong, session, consistent prefix
done
2.0
Segment compaction
sunby
in progress
2.0
Implements dynamic load balancing
sunby, xige-16
in progress
2.0
Implements dynamic handoff
xige-16, bigsheeper
in progress
2.0
Calculate distance between embeddings
done
2.1
Multi-tenant support and access control
pending
2.2
Change data capture
pending
Long Term
Adopts incremental backup
pending
Long Term
Supports static data encryption
pending
Long Term
Offers embedding-as-service through data importer/transformer
pending
Performance/Cost
Version
Feature
Owner
Status
Comment
2.0
Milvus 2.0 performance benchmark and tuning
czs007, dragondriver
in progress
2.1
Supports GPU Index building and embedding retrieval
shengjun1985
pending
2.1
Data bulkload
pending
2.1
Adopts cost-based query optimization algorithm to improve hybrid search efficiency
pending
2.1
Supports ScaNN Index
pending
2.2
Supports on-disk vector indexing
pending
Long Term
Supports FPGA and other Heterogeneous hardware
pending
Long Term
Automatic index optimization
pending
Stability
Version
Feature
Owner
Status
Comment
2.0.0-RC
Fully managed failure recovery and service discovery
done
2.0.0-RC
Python SDK test
done
2.0
Chaos test
yanliang567
pending
2.0
Pressure test
del-zhenwu
pending
2.1
Supports segment in memory replicas
pending
2.1
Flow control && back pressure support
pending
2.2
Query node resource isolation
pending
Ease Of Use
Version
Feature
Owner
Status
Comment
2.0.0-RC
Helm installation
done
2.0.0-RC
Support of Attu, a Milvus visual management tool
in progress
2.0
Prometheus, Grafana and Jaeger support
zwd1208
in progress
2.0
Milvus Kubernetes operator
zwd1208, jeffoverflow
pending
2.1
Multi datacenter deployment and multi-cloud integration
pending
2.2
Embedded Milvus that runs on laptops
pending
Long Term
Dynamic cluster expansion/shrink
pending
SDK
Version
Feature
Owner
Status
Comment
2.0.0-RC
Python ORM-style APIs
done
2.0
Merge PyMilvus ORM and PyMilvus
done
2.0
Supports NodeJs APIs
done
2.0
Supports Java SDK
xiaofan-luan
in progress
2.0
Supports Go SDK
congqixia
in progress
2.1
Supports RESTful APIs
pending
2.1
Supports C++ SDK
pending
Long Term
SQL-like query language
pending
Integration
Version
Feature
Owner
Status
Comment
2.0
Integrates S3
done
2.1
Integrates Kafka
pending
2.1
Integrates JuiceFS
pending
2.1
Data stored over local/distributed filesystems
pending
2.2
Integrates distributed KV stores such as HBase/TiKV/FoundationDB
