OceanBase Database Enterprise Edition V3.2.0|OceanBase Database|docs|OceanBase Documentations

Version information

Release date: July 5, 2021
Version: V3.2.0
RPM version: oceanbase-3.2.0-20210705162520

Overview

OceanBase Database V3.2.0 is a beta version for internal developers at OceanBase to verify and test the Hybrid Transactional and Analytical Processing (HTAP) capability of OceanBase Database.

Performance

A vectorization engine is supported to vectorize basic operators. The rewrite feature of the optimizer is adapted and enhanced. The goal is to improve HTAP capability based on the TPC-H data models.
The performance of the Nested Loop Join (NLJ) operator is improved by optimizing the data access cost for full table scans on small tables and the iterator reuse rules.
The data access process is optimized to reduce the instruction overhead of DML statements during the buffer write process.
The pre-read logic for accessing data rows is optimized to reduce the pre-read frequency and the amount of invalid data.
The performance of the GROUP BY operator is optimized. When the DISTINCT clause is specified, the computation is pushed down for local execution of the data. After that, the aggregated result is returned.
The compilation is optimized. The high-profile functions are rearranged during compilation to improve the hit rate of I-Cache.

Enhanced features

The UTF8 and GBK character sets support full-width characters in Oracle-compatible mode.
In Oracle-compatible mode, the following functions are added to the DBMS_STATS package to import and export the statistics: EXPORT_TABLE_STATS(), EXPORT_COLUMN_STATS (), IMPORT_TABLE_STATS(), and IMPORT_COLUMN_STATS().
In Oracle-compatible mode, the following functions are added to the DBMS_STATS package to lock and unlock the statistics: LOCK_TABLE_STATS (), LOCK_PARTITION_STATS (), LOCK_SCHEMA_STATS(), UNLOCK_TABLE_STATS (), UNLOCK_PARTITION_STATS (), and UNLOCK_SCHEMA_STATS().
Small files are optimized to support multiple versions. When an OBServer node is running, it generates and depends on some small files, such as cluster backup information files. When these files are backed up to OSS, concurrent access issues may occur when possible modifications are performed by the RootServer. Therefore, a multi-version file management mechanism is introduced, where the original individual files are organized into a directory that contains multiple files. Each of these files is identified by a specific version. The version number increases when the RootServer is switched.
In Oracle-compatible mode, you cannot create a tenant that uses the UTF16 character set.