Table 1 Features and documents related to 23.0.RC5
Feature
Document
Description
OmniData 1.5.0
OmniOperator 1.3.0
OmniShuffle 1.3.0
OmniRuntime Release Notes
OmniData documents are incorporated into the Kunpeng BoostKit 23.0.RC5 Big Data OmniRuntime Feature Guide. The OmniData feature accelerates native computing using the Filter operator and filters Parquet type data using the column index and a bloom filter. It supports overlay of operator pushdown and operator acceleration features on the Spark engine, the OmniVec data processing format, selectivity calculation for the string type, and pushdown of some tasks.
The OmniOperator feature improves the TPC-DS 99 query performance by 30%. TopNSort operators and Sort-Merge Join and Sort integration are added to implement the table scan native processing for Parquet files, and security clusters for ORC and Parquet.
The OmniShuffle feature is added to enable in-memory shuffle by implementing the Shuffle Manager plugin interface.
