Features of the serverless orchestration standard Knative and the workflow for using ACK Knative - Container Service for Kubernetes

Run serverless workloads on Kubernetes with request-based auto scaling, scale-to-zero, and canary releases. Fully compatible with community Knative and Kubernetes APIs, ACK Knative adds reserved instances and AHPA predictive scaling.

Why use Knative in a Kubernetes cluster

Knative

Knative is a Kubernetes-based framework that defines a cloud-native, cross-platform orchestration standard for serverless applications through container builds, workload management, and event-driven models. Benefits include:

Focus on business logic: Knative simplifies application configuration and auto scaling so developers can focus on business logic instead of underlying resources.
Standardization: Deploying serverless applications requires compiling source code, managing deployments, and handling events. Community and vendor solutions often lack a unified standard. Knative provides a universal framework for serverless applications.

For example, implement an event-driven architecture by writing and deploying a YAML custom resource (CR) in your cluster. This avoids deep integration with specific cloud products and simplifies cross-platform migration.
Ease of use: Knative packages code into container images, publishes them as services, and deploys functions to a Kubernetes cluster as containers.
Auto scaling and version management: Knative scales instances to zero when there is no traffic to save resources and supports version management and canary releases.
Event-driven: Knative provides a complete event model to ingest events from external systems and route them to services or functions.

The Introduction to the Knative application model describes the Knative Service application model.

Core components

Knative includes these core components:

Knative Serving: Manages serverless workloads with application deployment, multi-version management, request-based auto scaling, canary releases, and scale-to-zero when there is no traffic.
Knative Eventing: Integrates event sources and supports event registration, subscription, and filtering. Its event model decouples producers from consumers.
Knative Functions: Creates, builds, and deploys stateless, event-driven functions as Knative services on Kubernetes without deep knowledge of Kubernetes, containers, or Knative.

Why use ACK Knative

ACK Knative is fully compatible with community Knative and standard Kubernetes APIs, with enhanced product capabilities and broader solutions.

Enterprise-ready capabilities: One-click deployment eliminates the need to purchase resources or build the system yourself. The ACK console GUI simplifies Kubernetes and Knative operations.
Simplified operations:
- Managed core components: ACK creates and manages Knative Serving and Knative Eventing in ACK clusters for high availability at no charge for component resources.
- Managed gateways: ACK Knative offers Application Load Balancer (ALB), Service Mesh (ASM), and Kourier gateways. Except for community-compatible Kourier, ACK manages controllers for the other gateways as a fully managed, maintenance-free service.
Ecosystem integration: Integrates with Alibaba Cloud compute (ECI, ECS, and ACS), observability (Simple Log Service (SLS) and Prometheus), and application integration (EventBridge) so you can add logging, monitoring, alerting, continuous delivery, and event-driven capabilities to Knative services without provisioning servers or building services yourself.
Enhanced features: ACK Knative extends open source Knative with out-of-the-box solutions for real-world scenarios, such as:
- Reserved instances: Keep a low-cost, always-on instance for latency-sensitive applications to reduce cold start latency from scale-to-zero while controlling costs.
- Knative auto scaling: Provides request-based auto scaling with Knative Pod Autoscaler (KPA) and Horizontal Pod Autoscaler (HPA). Configure Advanced Horizontal Pod Autoscaler (AHPA) for cyclical workloads to provision resources in advance and reduce cold starts.

See the Comparison between ACK Knative and open source Knative for differences between ACK Knative and open source Knative.

Use cases

Typical ACK Knative use cases:

Scenario	Description
Web service hosting	Simplified deployment: ACK Knative abstracts low-level Kubernetes details and simplifies workload deployment through Knative services. Simplified multi-version management: The revision mechanism gives each revision a unique identifier for version management and rollbacks. Simplified traffic splitting for canary releases: ACK Knative assigns traffic percentages to revision services for canary releases, A/B tests, and other deployment strategies.
Serverless applications	Focus on business logic: Developers focus on business logic instead of IaaS resources, with simplified application configuration and lower infrastructure operational costs. On-demand resource usage and auto scaling: ACK Knative scales resources based on traffic and concurrency and scales instances to zero when there is no traffic.
AI workloads	Focus on business logic: In GPU and other heterogeneous computing scenarios, developers build and deploy AI tasks without maintaining underlying infrastructure. On-demand resource usage and auto scaling: ACK Knative scales resources based on workload to reduce costs for inference services with fluctuating demand. Portability: ACK Knative runs in any Kubernetes-compatible environment—cloud, on-premises data centers, or edge devices.
Event-driven workloads	Knative Eventing ingests events from external systems. For example, IoT devices can send sensor data to a Knative service. Configure ACK Knative event sources to receive data and trigger storage, real-time analysis, or monitoring and alerting.

Workflow

Workflow for using ACK Knative:

Step	Description
Prerequisites	An ACK managed cluster of version 1.22 or later. To upgrade your cluster, see Manually upgrade a cluster.
	Deploy ACK Knative and install Knative Serving per Deploy and manage Knative components.
	Select and deploy a gateway per Select a Knative gateway. ALB: A fully managed gateway based on Alibaba Cloud ALB. Provides high-capacity traffic management with automatic scaling and no O&M overhead. ASM: A managed, Istio-compatible platform for unified traffic management. Supports traffic shaping, mesh observability, and secure inter-service communication. Helps you manage services that run on heterogeneous computing infrastructure. Kourier: A lightweight, open source gateway from the Knative community, built on Envoy. Provides essential routing and service discovery.
Service deployment and management	Specify the resource type: ECS resources run Knative services by default. Use ECI resources in Knative to run ECI pods for traffic bursts. Use ACS resources to run ACS pods for traffic bursts. Use GPUs in Knative for scenarios such as AI inference services. If both ECS and ECI resources exist in the cluster, declare scaling order with a ResourcePolicy per Use both ECS and ECI resources in Knative. Use preemptible instances to reduce cloud computing costs. Configure a reserved instance to keep a low-specification burstable instance active and balance cost with startup time.
	Auto scaling: Implement KPA for a service based on the number of traffic requests to auto scale based on request rate (QPS). Use Knative and AHPA to implement scheduled auto scaling for predictive scaling from historical metrics and scheduled scaling with Cron expressions. Use HPA in Knative to scale based on CPU metric thresholds.
	Version management and canary releases: Create a revision for version management, such as rollbacks. Perform a canary release based on traffic splitting by assigning traffic percentages to revisions.
	Access to Knative services: The default Knative service domain is `{route}.{namespace}.{default-example.com}`, where `{default-example.com}` is the default suffix. Configure custom domains and paths to customize the suffix. For custom domains, configure a certificate to access services over HTTPS to secure data transmission. Configure port probing in Knative with Liveness Probe and Readiness Probe to monitor service health and availability.
Advanced features	Event-driven: See Knative Eventing for the serverless event model, including external event sources, routing, subscription, and filtering.
	Knative Functions: Deploy Knative Functions to create, deploy, and call functions in a Kubernetes cluster.
	AI inference services: Quickly deploy an inference service based on KServe to serve trained models (such as TFServing, TorchServe, and Triton) on a model serving runtime. ACK provides best practices for AI model inference in Knative, such as accelerating model deployment, and configuring GPU sharing scheduling. See Best practices for deploying AI model inference services in Knative.
	Service mesh: Integrate Service Mesh (ASM) with Knative services for traffic management and service security.
Observability and cost management	Log collection: Collect logs from Knative services with SLS to collect, consume, deliver, query, and analyze log data non-intrusively.
	Monitoring dashboard: View the Knative service monitoring dashboard in Managed Service for Prometheus for response latency, request concurrency, and more.
	Monitoring and alerting: Configure alerting for Knative services with SLS log-based alert rules.
	Enable the cost insights feature for Knative services to view resource usage and cost distribution.

Billing

When you use ACK Knative in an ACK cluster, no management fees are charged for ACK Knative. You are billed for created and used resources, such as cluster management fees, cloud servers, load balancer instances, and NAT gateways, based on their respective prices. See the Billing overview for ACK clusters and Cloud product resource fees.

FAQ

For ACK Knative issues, see the Knative FAQ.

For Knative questions, click here to join an organization.

References

Upgrade Knative Serving promptly for the latest features and bug fixes. See Knative Release Notes and Upgrade Knative Serving component.
The official Knative Knative Bookstore Tutorial walks through building, deploying, and monitoring a bookstore application.