Kylin 4.0 Performance Tuning Guide: Optimizing Cube Build and Query

Kylin 4.0 represents a major architectural upgrade compared with previous versions, featuring a brand‑new build engine and query engine that greatly improve build and query performance, address single‑point issues, remove HBase dependency, and make compute‑storage separation and cloud‑native deployment possible.

With the release of the Kylin 4.0‑alpha version, the tuning methods have changed significantly, so most of the tuning experience accumulated from earlier versions no longer applies. To help users quickly understand the new tuning approach and avoid performance bottlenecks, Kyligence’s big‑data architect Zhang Zhichao published a performance‑tuning guide for Kylin 4.0.

A B‑station video was also released, which is divided into two parts: “Cube Build Performance Optimization” and “Query Performance Optimization”.

The “Cube Build Performance Optimization” part explains how, beyond the traditional dimension hierarchy and required‑dimension pruning, users can improve build speed by configuring appropriate Spark resources, leveraging Kylin 4.0’s automatic parameter tuning, and adjusting global dictionary and dimension‑table snapshot settings.

The “Query Performance Optimization” part describes techniques such as specifying shardBy columns, setting the sortBy order, reducing small files, and tuning the configuration of the query engine SparderContext .

The guide’s directory is illustrated in the image below:

Additional recommended readings are provided:

Kylin 在贝壳找房的实践及 HBase 优化

网易云音乐基于 Flink + Kafka 的实时数仓建设实践

分布式存储引擎 Alluxio 入门指南