Alibaba Cloud Container Compute Service (ACS) is a container service product based on Kubernetes, providing compute resources compliant with container specifications. The delivery model of ACS compute power is Serverless, which means users do not need to be concerned with the operation and management of nodes and clusters. ACS can support your container applications and cloud products in varieties scenarios.
What is ACS?
ACS enhances the Kubernetes capabilities and optimizes the orchestration and scheduling flexibility of Serverless Kubernetes with good user experience based on the ACK Serverless cluster. ACS clusters are also a type of Serverless Kubernetes cluster in ACK. You can view related ACS clusters on the ACK console and manage role-based access control (RBAC) authorization and kubeconfig credentials for ACS clusters. ACS defines four container compute options: three high-cost-performance CPU container computes and one GPU container compute, tailored for container business scenarios.
Compute type - general-purpose | Compute type - performance | Compute type - GPU | |
QoS type - default | √ | √ | √ |
QoS type - best-effort | √ | × | × |
GPU container compute power is in invitational preview. If you need to use or experience it, submit a ticket.
Benefits
Quick to start
You can use general-purpose or performance container compute power for online business scenarios. For offline business scenarios, cost-effective BestEffort container compute power is available. You can quickly create and deploy business through simple configuration with YAML files or the console.
Flexible application
Container compute power starts at a minimum specifications of 0.25 vCPU and 0.5 GiB, and increments by 0.5 vCPU and 1 GiB. You can flexibly apply for and use resources as needed to reduce resource costs.
Pay-as-you-go
ACS supports large-scale, second-level elasticity. You can flexibly apply for and use resources as needed. ACS is billed on a pay-as-you-go basis, and you can also use daily consumption plans which allow for more economical and flexible response to fluctuations in business traffic.
Rich Scenarios
ACS offers a variety of business scenarios and can support your container applications and cloud products. Core system components of Kubernetes clusters are managed by ACS. ACS also supports automatic upgrades of cluster patch versions which significantly reduces the complexity of cluster maintenance.
Scenarios
General online business
You can use general-purpose instances with high cost-performance, low-fluctuation, and performance-stable in common business scenarios for microservices applications, web applications, and network servers. ACS supports second-level startup and large-scale elastic scaling to ensure stable handling of business requests with sudden traffic spikes, and also avoid traffic loss or cost waste caused by premature scaling operations.
Big data computing business
You can use high cost-performance BestEffort instances in business scenarios that are not sensitive to latency but focus on data processing throughput, such as Spark, Presto, AI training. ACS supports second-level startup and large-scale elastic scaling to reduce task backlogs leading to queue wait times and improve the efficiency of large-scale parallel computing business.
AI training and inference business
You can combine general-purpose GPU computing power and GPU capacity reservation in business scenarios that are latency-sensitive real-time deployment of AI inference, such as AI-Generated Content (AIGC) model training and inference, autonomous driving training and inference, and real-time inference tasks for cloud graphics. This approach provides the most cost-effective solution while ensuring resource determinism.
High-performance business
You can use ACS in business scenarios with higher performance requirements, such as cloud gaming. ACS supports second-level startup and large-scale elastic scaling, ensuring stable handling of business requests with sudden traffic spikes. This provides a smooth user experience and avoids delays or lag due to insufficient processing capacity.
Core features
Resource management
Feature item | Description |
Instance type | ACS provides three different pod instance types, general-purpose, performance, and GPU for different business scenarios. For more information, see ACS pod overview. |
On-demand elasticity | ACS uses on-demand elastic resources by default, you can apply for resources based on different instance types of pods. ACS is billed on a second-level pay-as-you-go basis. You can view the corresponding metering data through the bill details. |
Capacity reservation | ACS is designed for the business scenarios of GPU inference and provides the capability of capacity reservation for pod. |
Cluster management
Feature item | Description |
Cluster creation | ACS clusters and resources are integrated, you do not need to manage clusters and nodes. You can create and deploy business through simple configuration with YAML files or the console, which significantly lowers the threshold for using Kubernetes and cloud. For more information, see Create an ACS cluster. |
Connect to cluster | You can connect to the cluster by obtaining kubeconfig and using the kubectl tool, you can manage and use ACS clusters and resources in the terminal. For more information, see Obtain cluster KubeConfig and connect to the cluster using the kubectl tool. |
Authorization management | ACS supports Alibaba Cloud Resource Access Management (RAM) authorization and Role Based Access Control (RBAC) authorization management of the Kubernetes cluster. For more information, see Authorization overview. |
Application management
Feature item | Description |
Application creation | ACS supports various types of workloads, including stateless, stateful, task, and scheduled tasks. You can create an application from clients, console images, and templates, and config environment variables, application health, data disk, and log. For more information, see Create stateless workload deployment. |
Application scaling | ACS supports manual scaling, Horizontal Pod Autoscaling (HPA) automatic scaling policies, and Cron Horizontal Pod Autoscaler (CronHPA) automatic scaling policies. For more information, see Implement horizontal pod autoscaling and CronHPA. |
Storage management | ACS integrates Alibaba Cloud storage services, including cloud disk Elastic Block Storage (EBS) and File Storage NAS (NAS), based on the Kubernetes Container Storage Interface (CSI). For more information, see Storage CSI overview. |
Network management | ACS integrates Kubernetes networking, Alibaba Cloud Virtual Private Cloud (VPC), and Alibaba Cloud Server Load Balancer (SLB), which together provide a stable and high-performance container network. For more information, see Network overview. |
O&M and security
Feature category | Sub-feature | Description |
Observability | Monitoring | ACS integrates and enables Alibaba Cloud Prometheus by default, which determines the monitoring status of Kubernetes clusters, instance pods, and applications across multiple dimensions. This determination is based on preset monitoring dashboards and performance metrics. For more information, see Use Alibaba Cloud Prometheus to monitor ACS cluster status. |
Logs | ACS integrates Simple Log Service (SLS). You can collect and view application logs, pod instance logs, and cluster logs. For more information, see Configure application log collection through pod environment variables. | |
Alerts | ACS supports alerts for abnormal events in container services and container scenario metrics. | |
Cluster inspection | Cluster inspection | ACS provides inspection functionality. You can automatically scan cluster health state to discover potential risks within the cluster. For more information, see Use cluster inspection to identify potential threats in the cluster. |
Fault diagnosis | ACS provides one-click fault diagnosis capabilities, which include pod, service, and Ingress diagnostics. For more information, see Use cluster diagnosis. | |
Security center | Audit | ACS provides audit log functionality for API Server, which helps cluster administrators to record or trace daily operations of different users. For more information, see Work with cluster auditing. |
More Information
For the latest updates of ACS, see Release notes.