TiDB RC3 Release Notes

On June 20, 2017, TiDB RC4 is released!This release is focused on MySQL compatibility, SQL optimization, stability, and performance.

Highlight

• The privilege management is refined to enable users to manage the data access privileges using the same way as in MySQL.
• DDL is accelerated.
• The load balancing policy and process are optimized for performance.
• TiDB Ansible is open sourced. By using TiDB-Ansilbe, you can deploy, upgrade, start and shutdown a TiDB cluster with one click.

Detailed updates

TiDB

• The following features are added or improved in the SQL query optimizer:
  • Support incremental statistics
  • Support the Merge Sort Join operator
  • Support the Index Lookup Join operator
  • Support the Optimizer Hint Syntax
  • Optimize the memory consumption of the Scan, Join, Aggregation operators
  • Optimize the Cost Based Optimizer (CBO) framework
  • Refactor Expression
• Support more complete privilege management
• DDL acceleration
• Support using HTTP API to get the data distribution information of tables
• Support using system variables to control the query concurrency
• Add more MySQL built-in functions
• Support using system variables to automatically split a big transaction into smaller ones to commit

Placement Driver (PD)

• Support gRPC
• Provide the Disaster Recovery Toolkit
• Use Garbage Collection to clear stale data automatically
• Support more efficient data balance
• Support hot Region scheduling to enable load balancing and speed up the data importing
• Performance
  • Accelerate getting Client TSO
  • Improve the efficiency of Region Heartbeat processing
• Improve the pd-ctl function
  • Update the Replica configuration dynamically
  • Get the Timestamp Oracle (TSO)
  • Use ID to get the Region information

TiKV

• Support gRPC
• Support the Sorted String Table (SST) format snapshot to improve the load balancing speed of a cluster
• Support using the Heap Profile to uncover memory leaks
• Support Streaming SIMD Extensions (SSE) and speed up the CRC32 calculation
• Accelerate transferring leader for faster load balancing
• Use Batch Apply to reduce CPU usage and improve the write performance
• Support parallel Prewrite to improve the transaction write speed
• Optimize the scheduling of the coprocessor thread pool to reduce the impact of big queries on point get
• The new Loader supports data importing at the table level, as well as splitting a big table into smaller logical blocks to import concurrently to improve the data importing speed.