pingcap/tikv
Tikv
TiKV is a cloud-native distributed transactional key-value store and storage engine. It provides a distributed database designed for horizontal scalability and strong consistency across a cluster of physical nodes.
The system uses a Raft-based consensus mechanism to maintain data availability and state synchronization. It ensures ACID compliance for distributed transactions through a two-phase commit workflow and manages data distribution via multi-Raft sharding.
The engine handles massive datasets using automated range splitting and cluster load balancing to distribute data across different physical zones or geographic regions. To reduce network latency, it employs a coprocessor for distributed computing, allowing filtering and aggregation logic to execute directly on storage nodes. Local data persistence is managed via a log-structured merge-tree engine.
Features
- Distributed Key-Value Stores - Provides a consistent and replicated distributed key-value store designed to maintain a reliable source of truth across a cluster.
- Distributed Consensus Stores - Provides a distributed transactional key-value store that maintains consistent state across nodes using Raft consensus.
- Data Replication Strategies - Copies data across multiple nodes in consensus groups to ensure operational continuity and fault tolerance during node failures.
- Cloud-Native Databases - Engineered as a cloud-native database with support for geographic data distribution and seamless horizontal scaling.
- Multi-Raft Sharding - Divides the keyspace into small ranges, each managed by its own independent Raft consensus group for horizontal scalability.
- Consensus Algorithms - Maintains a consistent view of data across all nodes using a Raft-based consensus algorithm to prevent state divergence.
- Distributed Storage Engines - Provides a storage system that shards data across multiple physical nodes with automated balancing and replication.
- Distributed Transaction Processing - Stores key-value data using a distributed layer that guarantees ACID compliance for all transactional operations via consensus protocols.
- Horizontal Scaling - Distributes data across multiple nodes using automated sharding and migration to scale storage and throughput horizontally.
- Log-Structured Merge-Trees - Uses a log-structured merge-tree engine for high-throughput local data persistence on individual nodes.
- Raft Consensus Implementations - Maintains data consistency and availability by synchronizing a replicated log across storage nodes using the Raft protocol.
- Two-Phase Commit Protocols - Ensures ACID compliance across distributed shards using a two-phase commit coordinator to manage prepare and commit workflows.
- Distributed Transaction Management - Handles complex data operations across multiple nodes while ensuring ACID compliance and strict data consistency.
- Dynamic Shard Splitting - Automatically divides oversized data shards into smaller pieces to balance load and manage storage growth.
- Cluster Load Balancing - Automatically redistributes data ranges across the cluster to balance storage and compute load across physical nodes.
- Cross-Zone Data Distribution - Distributes data copies across specific physical zones or geographies to ensure high availability and disaster recovery.
- Coprocessors - Implements a coprocessor for executing filtering and aggregation logic directly on storage nodes to minimize network latency.
- Distributed Computing - Executes filtering and aggregation logic directly on storage nodes to reduce data transfer over the network.
- Regional Replication - Spreads data copies across different physical locations to ensure high availability and protection against regional disasters.
- Storage-Side Computation - Processes data directly on storage nodes using a coprocessor to reduce network traffic and lower system latency.
- Key-Value Stores - Distributed key-value database powered by Rust.
