TiDB RC4 Release Notes

On August 4, 2017, TiDB RC4 is released! This release is focused on MySQL compatibility, SQL optimization, stability, and performance.

Highlight

• For performance, the write performance is improved significantly, and the computing task scheduling supports prioritizing to avoid the impact of OLAP on OLTP.
• The optimizer is revised for a more accurate query cost estimating and for an automatic choice of the Join physical operator based on the cost.
• Many enhancements have been introduced to be more compatible with MySQL.
• TiSpark is now released to better support the OLAP business scenarios. You can now use Spark to access the data in TiKV.

Detailed updates

TiDB

• The SQL query optimizer refactoring:
  • Better support for TopN queries
  • Support the automatic choice of the of the Join physical operator based on the cost
  • Improved Projection Elimination
• The version check of schema is based on Table to avoid the impact of DDL on the ongoing transactions
• Support BatchIndexJoin
• Improve the Explain statement
• Improve the Index Scan performance
• Many enhancements have been introduced to be more compatible with MySQL
• Support the JSON type and operations
• Support the configuration of query prioritizing and isolation level

Placement Driver (PD)

• Support using PD to set the TiKV location labels
• Optimize the scheduler
  • PD is now supported to initialize the scheduling commands to TiKV.
  • Accelerate the response speed of the region heartbeat.
  • Optimize the balance algorithm
• Optimize data loading to speed up failover

TiKV

• Support the configuration of query prioritizing
• Support the RC isolation level
• Improve Jepsen test results and the stability
• Support Document Store
• Coprocessor now supports more pushdown functions
• Improve the performance and stability

TiSpark Beta Release

• Implement the prediction pushdown
• Implement the aggregation pushdown
• Implement range pruning
• Capable of running full set of TPC-H except one query that needs view support