Kubernetes를 활용한 서버리스(Knative)

Welcome to the world of serverless on Kubernetes! Serverless computing allows you to run code without provisioning or managing servers, focusing purely on your application logic.

By combining serverless with Kubernetes, you get the benefits of both: the agility and cost-efficiency of serverless, backed by the robust, scalable, and portable infrastructure management of Kubernetes.

Knative is an open-source platform that extends Kubernetes to provide components for deploying, running, and managing serverless workloads. It simplifies the complexities of running serverless applications.

Knative has two main components: Knative Serving, which handles request-driven services, and Knative Eventing, for event-driven architectures.

Knative Serving is designed for deploying and managing stateless services. It takes care of:

• Rapid Deployment: Quickly deploys container images.
• Auto-scaling: Scales services up and down based on demand, even to zero instances when idle.
• Revision Management: Provides immutable revisions for easy rollbacks and traffic splitting.

It's perfect for web applications, APIs, and microservices.

Key concepts in Knative Serving:

• Service: The top-level resource, defining the desired state of your application.
• Revision: An immutable snapshot of your service's code and configuration at a specific point in time. New deployments create new revisions.
• Route: Manages network traffic to one or more revisions, enabling features like A/B testing or gradual rollouts.

These components work together to provide a robust deployment and scaling mechanism.

A Knative Service is defined using a YAML manifest, much like other Kubernetes resources. This example defines a simple 'hello-knative' service using a pre-built image.

Notice the serving.knative.dev/v1 API version and the kind: Service.

This is a basic Python Flask application that Knative would deploy. It listens for HTTP requests and responds with a greeting, using an environment variable for customization. Knative handles the containerization, deployment, and scaling of this app.

Knative Eventing enables event-driven architectures on Kubernetes. It allows services to communicate asynchronously by sending and receiving events, decoupling the event producers from consumers.

This is crucial for building reactive systems that respond to various changes, like data updates, file uploads, or scheduled tasks.

Key components in Knative Eventing:

• Event Source: Generates events from various systems (e.g., Kafka, GitHub, CronJob, S3).
• Broker: A central messaging hub where event sources send events. It abstracts the underlying messaging system.
• Trigger: Filters events from a Broker based on attributes and routes them to a specific event consumer (often a Knative Service).

This flow enables flexible and scalable event processing.

Imagine you have a service that processes new images. You can set up an Event Source that monitors an S3 bucket for new image uploads.

When an image is uploaded, the source sends an event to a Broker. A Trigger, configured to filter for 'new S3 image' events, then forwards that event to your image processing Knative Service.

Your service processes the image, scales down when idle, and you only pay for the compute used during processing.

Which Knative component is primarily responsible for automatically scaling services up or down based on incoming requests, including scaling to zero when idle?

In this lesson, you explored how Knative brings serverless capabilities to Kubernetes. You learned about:

• Knative Serving: For deploying and auto-scaling request-driven services.
• Knative Eventing: For building event-driven architectures with sources, brokers, and triggers.

Knative empowers developers to build and run serverless applications with the control and flexibility of Kubernetes.