OmniRuntime Overview
OmniRuntime consists of a series of features provided by the Kunpeng BoostKit for Big Data in terms of application acceleration. It aims to improve the performance of end-to-end data loading, computing, and exchange through plugins, thereby improving the performance of big data analysis.
OmniRuntime includes the following features: operator pushdown (OmniData), operator acceleration (OmniOperator), shuffle acceleration, and materialized view (OmniMV). In the data loading phase, OmniData can be used to implement near-data computing, reducing network data traffic. In the data computing phase, OmniOperator can be used to replace native Java operators with high-performance native operators to improve operator efficiency. In the data exchange phase, the shuffle acceleration feature can be used to accelerate data interaction between nodes, which will be released in a later version. In addition, for scenarios where repeated queries or subqueries exist, OmniRuntime provides OmniMV to identify the optimal materialized view through AI algorithms, reducing the overhead of repeated subqueries and thus improving query efficiency.
