Network Planning

A storage and compute coupled network is recommended for OmniScheduler. Use nodes as both storage and compute nodes to maximize the load balancing effect in big data scenarios.

This storage and compute coupled network consists of four servers, which are one management node and three compute nodes. In the following, we will be using HDFS on the storage nodes for illustration:

• The management node is server1, which is used to manage tasks and run Prometheus, Grafana, and LoadsMetric.
• The compute nodes are agent1, agent2, and agent3, which are used to run compute services and Node Exporter.