Deployment Models

This section describes two deployment models of on-premises clusters: HA and non-HA.

HA deployment model

In this model, an on-premises cluster requires three master nodes, which improve cluster reliability and fault tolerance. Even if one master node becomes faulty, the cluster can still run normally. For environments where production services are running, use this model to prevent business loss caused by individual faults.

Non-HA deployment model

In this model, an on-premises cluster requires only one master node. If this node becomes faulty, the cluster cannot run normally. This model is not a good choice for the production environment. In this model, you can run workloads on the master node. It is suitable for non-critical or resource-constrained service scenarios, such as development and testing environments and edge environments. The cluster computing requires as low as 4 vCPUs and 8-GiB memory, which significantly lowers system resource requirements.