MySQL Pluggable Online Vectorized Analysis Engine

Application scenario: The MySQL Pluggable Online Vectorized Analysis Engine (KOVAE for short) applies to OLAP query, large data volume, and multi-core CPU scenarios.

Technical principle: This feature adopts a parallel processing mechanism. Each operator contains multiple worker threads, which significantly shortens the execution time of SQL statements. The columnstore structure improves data processing efficiency, and a secondary execution engine is provided as a plugin which is pluggable and can be dynamically loaded.

Figure 1 shows the data flow of SQL statements and result sets after execution.

Figure 1 KOVAE working principle

Figure 2 shows the KOVAE internal execution process.

Figure 2 Parallel execution architecture

Performance metric: The concurrent acceleration technology provided by KOVAE improves the query performance of OLAP by over three times.

**Table 1** SQL operators that support parallel query
Operator	Description
TableScanIterator	Full table scan operator iterator
IndexRangeScanIterator	Index range scan iterator
RefIterator	Operator iterator of equality matching on non-unique indexes
EQRefIterator	Equality matching on the unique index
LimitOffsetIterator	Limit operator
FilterIterator	Filter operator
NestedLoopIterator	NestedLoopJoin operator
HashJoinIterator	HashJoin operator
AggregateIterator	Aggregate operator
SortingIterator	Sorting operator
TemptableAggregateIterator	Aggregate operator for saving grouping results in a temporary table
MaterializeIterator	Materialize operator
StreamingIterator	Streaming operator

Restrictions:

Currently, KOVAE adapts only to MySQL 8.0.25. Other MySQL versions are not verified.
KOVAE performs parallel executions. If the SQL statements specify no result set sorting information, the result set may be sorted differently from that of the open source MySQL.
Only some SQL statements can be offloaded to KOVAE for execution. For details, see the MySQL Pluggable Online Vectorized Analysis Engine Feature Guide. You can add hints or set cost thresholds to control the offloading conditions for SQL statements.

Parent topic: Features