You can register clusters that are deployed in data centers or on a third-party cloud to Alibaba Cloud Distributed Cloud Container Platform (ACK One). This allows you to colocate clusters in ACK One and manage the clusters in a centralized manner. This topic describes the features and use scenarios of registered clusters.
Entry point for registered clusters in the ACK One console
Features
Your daily operations and maintenance work may involve Container Service for Kubernetes (ACK) clusters, self-managed Kubernetes clusters deployed in data centers, and Kubernetes clusters deployed on third-party clouds. If you have the following requirements for managing multiple clusters that are deployed in a multi-cloud or a hybrid cloud environment, we recommend that you register the clusters to ACK One.
Hybrid cloud elasticity: You want to add VMs, physical servers, and serverless computing resources in the cloud to registered clusters. ACK One provides multiple scaling and scheduling policies based on ack-co-scheduler. You can use these policies to scale out resources based on priorities and scale in resources on demand in the cloud and data centers. You can also use these policies to manage the ratio of replicated pods in the cloud and data centers, and scale node pools that run GPU-accelerated training jobs on demand.
Centralized management of ACK clusters and self-managed Kubernetes clusters deployed in data centers: The ACK console provides consistent O&M experience and security governance capabilities that allow you to centrally manage ACK clusters and self-managed Kubernetes clusters deployed in data centers. For example, the ACK console provides a unified logging, monitoring, and alerting system, and supports unified authorization management of Alibaba Cloud accounts, Resource Access Management (RAM) users, and RAM roles. This allows you to centrally manage all your clusters and applications.
AI and big data capabilities: ACK One provides topology-aware CPU scheduling and enables non-uniform memory access (NUMA) awareness to improve computing efficiency by 30% to 40%. ACK One supports GPU sharing and scheduling to improve GPU utilization by 300%. ACK One allows you to scale and centrally manage heterogeneous resources in the cloud and data centers. ACK One is integrated with Fluid to provide unified and accelerated access to distributed cache in a hybrid cloud environment. Fluid accelerates data access by about 10 times and reduces bandwidth usage by 90%.
Data backup and disaster recovery: ACK One allows you to back up, restore, and migrate data to the cloud. This ensures service continuity.
Usage scenarios
Scenario 1: Connect an external cluster in a data center to a registered cluster
Scenario
External clusters in data centers: ACK One allows you to connect a data center to the cloud and share resources.
On-demand resource and application scaling: ACK One can scale out resources and applications in the cloud during peak hours to balance the traffic load in the external cluster.
Scenario 2: Use registered clusters and Alibaba Cloud services to extend the capability of external clusters and provide consistent cloud and on-premises O&M
Scenario
Consistent cloud and on-premises O&M: You can use O&M capabilities provided by ACK One in Kubernetes clusters deployed in data centers or third-party public clouds.
Enhanced observability: ACK One allows you to collect logs, monitoring data, and events, and offers consistent cloud and on-premises O&M experience across different environments.
Enhanced security: You can enable auditing, security inspection, node risk detection, and policy governance with a few clicks.
Microservice governance: You can use ASM and MSE to perform microservice governance.
Scenario 3: Use registered clusters to implement disaster recovery in hybrid cloud environments, within a region, or across regions
Scenario
Deploy applications to the cloud: ACK One allows you to back up and restore applications across regions and from a data center to the cloud. This helps you deploy your business to the cloud in an efficient manner.
Disaster recovery: ACK One allows you to back up stateful applications across regions and from a data center to the cloud. ACK One also allows you to configure backup policies and restoration policies. This allows you to continuously back up your applications to the cloud to enhance the availability of your businesses.
Disaster recovery: ACK One provides scheduled data backup and geo-disaster recovery for applications and data in the cloud and data centers.
Active geo-redundancy: ACK One allows you to build a disaster recovery system by deploying three data centers across two zones. This feature is compatible with Kubernetes and helps ensure business continuity.
Scenario 4: Use registered clusters and the ACK scheduler to empower AI and big data
Scenario
AI algorithm development: ACK One allows you to manage AI jobs, quotas, and observability.
AI training: ACK One supports topology-aware scheduling and a variety of job scheduling policies. This allows you to improve AI training efficiency. ACK One supports compute-storage decoupling, which accelerates distributed training jobs. ACK One supports cross-cluster scheduling of various types of jobs, such as Spark jobs, Kubernetes jobs, and TensorFlow training jobs.
AI inference: ACK One supports GPU sharing, which can improve GPU utilization by approximately 300%. ACK One can automatically scale resources that are deployed on the cloud and in data centers.
Intelligent CPU scheduling: ACK One supports intelligent CPU scheduling and non-uniform memory access (NUMA) awareness based on ECS Bare Metal instances.