PolarDB for Oracle provides the cross-node parallel execution and uses an optimizer to ensure high performance.
Cross-node parallel execution of transactional and analytical queries
PolarDB for Oracle supports analytical queries. PolarDB for Oracle uses the distributed parallel computing technology to run an SQL query on multiple read-only nodes in parallel. This increases the I/O throughput at the shared storage layer and improves the CPU utilization and memory usage of read-only nodes. Analytical queries on transaction data can be run in real time without the need to import data to an analysis platform. The read-only nodes in which analytical queries are run are physically isolated from the read-only nodes in which transactional queries are run. This way, analytical queries on read-only nodes cannot affect transaction processing services. When computing resources are insufficient, compute nodes can be added on demand. The increase of compute nodes does not require data re-allocation.
The optimizer module is one of the most attractive modules in PolarDB for Oracle. The optimizer module consists of two key parts: the logical optimizer and the physical optimizer.
The logic optimizer optimizes operations such as query rewriting to improve execution plans or to increase the search space.
The physical optimizer optimizes connection sequences, connection algorithms, and access paths based on metadata and statistics. PolarDB for Oracle estimates the cost based on the model and then chooses the execution plan that costs the least to ensure that SQL statements are executed in an efficient manner.
Simple Permission Model (SPM) allows you to manage SQL statements in an online manner. You can change the execution plan for a query without the need to modify the SQL statements.