Resource Pool Federation Overview

Overview

Preparations and standard cluster creation related to a resource pool federation need to be performed on both the local cloud and collaboration cloud. The creation and management of a resource pool federation only need to be performed on the local cloud.

In a multi-cloud environment, compute resources are unbalanced in resource pools on different cloud platforms. The compute resources of a single resource pool or region may be insufficient to support large-scale job running, which limits service expansion and efficiency. To solve this problem, we provide the capability of establishing a federation across cloud resource pools. Resource pools of different cloud platforms can be federated to form a resource pool federation. Through the resource pool federation, jobs can be delivered to different resource pools to use compute across clouds, effectively solving the problem of insufficient compute in a single resource pool or region. Specifically, we have implemented A2&A3 heterogeneous hybrid cross-cloud federation training on HCSO and Huawei Cloud, further improving resource utilization and service flexibility.

Concepts

Resource pool federation: provides the capability of forming a federation across cloud resource pools. Resource pools on different cloud platforms can be federated to form a resource pool federation. Jobs can be delivered to different resource pools through the resource pool federation, enabling cross-cloud use of compute and addressing the issue of insufficient compute resources in a single resource pool or region. This enables A2&A3 heterogeneous hybrid cross-cloud federation training between HCSO and Huawei Cloud.

Operations

This section describes how to manage the lifecycle of a resource pool federation, connect networks across clouds, manage resource pools, and deliver training jobs to the resource pool federation on the training page.

• Local cloud: You need to create a standard cluster 2.0 and resource pool federation on the local cloud.
• Collaboration cloud: You need to create a standard cluster 2.0 on the collaboration cloud.