TiDB Roadmap

This document defines the roadmap for TiDB development.

TiDB:

  • Optimizer
    • Refactor Ranger
    • Optimize the cost model
    • Join Reorder
  • Statistics
    • Update statistics dynamically according to the query feedback
    • Analyze table automatically
    • Improve the accuracy of Row Count estimation
  • Executor
    • Push down the Projection operator to the Coprocessor
    • Improve the performance of the HashJoin operator
    • Parallel Operators
      • Projection
      • Aggregation
      • Sort
    • Compact Row Format to reduce memory usage
    • File Sort
  • View
  • Window Function
  • Common Table Expression
  • Table Partition
  • Cluster Index
  • Improve DDL
    • Speed up Add Index operation
    • Parallel DDL
  • Support utf8_general_ci collation

TiKV:

  • Raft
    • Region merge
    • Local read thread
    • Multi-thread raftstore
    • None voter
    • Pre-vote
    • Multi-thread apply pool
    • Split region in batch
    • Raft Engine
  • RocksDB
    • DeleteRange
    • BlobDB
  • Transaction
    • Optimize transaction conflicts
    • Distributed GC
  • Coprocessor
    • Streaming
  • Tool
    • Import distributed data
    • Export distributed data
    • Disaster Recovery
  • Flow control and degradation

PD:

  • Improve namespace
    • Different replication policies for different namespaces and tables
  • Decentralize scheduling table Regions
  • Scheduler supports prioritization to be more controllable
  • Use machine learning to optimize scheduling
  • Cluster Simulator

TiSpark:

  • Limit/Order push-down
  • Access through the DAG interface and deprecate the Select interface
  • Index Join and parallel merge join
  • Data Federation

SRE & tools:

  • Kubernetes based integration for the on-premise version
  • Dashboard UI for the on-premise version
  • The cluster backup and recovery tool
  • The data migration tool (Wormhole V2)
  • Security and system diagnosis