Documentation
Rate This Document
Findability
Accuracy
Completeness
Readability
SubmitClose

Tuning Guidelines

  1. The HiBench tuning cases use the Kafka component as the data source. Therefore, you need to increase the pressure on Kafka data production to ensure that the Kafka does not become a performance bottleneck.
  2. The single-node deployment mode is used in Storm component tuning, which reduces performance overheads caused by inter-node communication and serialization. Therefore, topology concurrency tuning is especially important.
  3. The statistics page of Storm is used as an important method to locate concurrency bottlenecks of the current topology. In addition, the quantity of pending and acker threads is also analyzed for optimal performance.
  4. This document applies to the HDP platform. You can refer to this document for the Apache platform.
Parent topic: Introduction