Kubernetes and Cloud-Native Applications

Kubernetes and Cloud-Native Applications

• Kubernetes is a container orchestration platform, initially developed by Google and released as an open-source project in 2014.
• It is a de facto standard container-as-a-service (CaaS) platform.
• Trajche Krstev is a Solution Architect for Packet Core and Telco Cloud.

Agenda

• Container as a Service (CaaS)
• Cloud evolution
• The path to cloud native
• Cloud Native Trail Map
• Containers background
• Kubernetes architecture
• Kubernetes overview

Container as a Service (CaaS)

• CaaS is a cloud computing model that manages, deploys, and scales containerized applications.
• It streamlines container lifecycle management (provisioning, scheduling, scaling, and monitoring).
• Key features include container management, orchestration support, flexibility across cloud environments (hybrid, public, private), and API-driven automation.
• Benefits include developer efficiency, portability, pay-as-you-go pricing, and rapid deployment.

The Cloud Evolution

• Legacy systems face challenges like hardware coupling, monolithic architectures, limited scaling, and low automation.
• IaaS systems offer hardware independence, but limited scaling and partial automation.
• Virtual Network Functions (VNF) systems have hardware independence, mostly monolithic, and limited scaling.
• Cloud native systems feature microservices architecture, high availability, continuous deployment, full automation, and complete hardware independence.
• CaaS platforms like Kubernetes offer a modern approach to cloud computing.

The Path to Cloud Native

• The Cloud Native Computing Foundation (CNCF) is the Linux Foundation community for containers and cloud-native applications.
• CNCF hosts open-source projects like Kubernetes, key to realizing cloud-native applications.
• The Cloud Native Landscape Project is a resource map for enterprises and developers transitioning to cloud-native technologies.
• CNCF projects represent a well-established path for cloud-native application deployment.

The Cloud Native Trail Map

• Addresses containerization, continuous integration/continuous delivery (CI/CD), orchestration and application definition, observability and analysis, service proxy, discovery, and mesh, networking, policy, security, distributed database, and storage, container registry, and runtime, and software distribution.
• This trail map assists in navigating the cloud-native journey.

Containers Background

• Containers provide a sandboxed execution environment for processes.
• Container images are stored in registries.
• Containers are lightweight, focusing on process-level CPU/memory usage.
• Filesystem layers are shared among containers.
• Communications are enabled through bridges or similar inter-process communication methods.
• Docker images are stored in a registry and can be retrieved by name and version.
• Containers run in a single server or can be orchestrated across multiple servers.

Kubernetes Architecture

• Kubernetes is an orchestration platform that manages containerized applications.
• Control plane: Master nodes for high availability (typically replicated). Includes components like API server, controller manager, scheduler, and etcd.
• Runtime plane: Worker nodes (minions). For deploying and managing Pods.
• Kubernetes architecture is comprised of the control plane (managing applications), worker nodes (executing tasks), Kubernetes API (to allow external access), and Container Runtime Interface (CRI).
• Kubectl refers to the command-line tool for interacting with Kubernetes. The kubeconfig file automatically discovers API endpoints and credentials.
• There are namespaces for light multitenancy and resource quota enforcement.

Kubernetes - Architecture (Control Plane)

• The API server interacts with clients and supports REST-based APIs. It acts as the gateway for all control plane components.
• Kubernetes uses etcd to store cluster configuration, runtime state, and metadata. This ensures data consistency and availability for the control plane.
• The scheduler assigns Pods to appropriate nodes according to various cluster policies.
• The controller manager orchestrates Kubernetes resources (nodes, workloads, configurations), continuously monitoring and maintaining appropriate cluster states.
• The cloud controller manager interacts with the underlying cloud and integrates with cloud providers.

Kubernetes - Architecture (User Plane)

• Kubelet is the control plane agent in every worker node, managing the execution of containers.
• Kube-proxy is responsible for network configurations for applications and services, including CoreDNS handling for DNS resolution.
• The container runtime manages container execution and interactions (e.g., Docker).
• Container Runtime Interface (CRI) governs container runtime specifications, interfaces, and APIs used by kubelet.
• Container Network Interface (CNI) provides a standard interface between the container runtime and networking.

Kubernetes Cluster, Pods, Deployments, and Services

• Kubernetes deployments manage Pods.
• Pods are the smallest deployable units in Kubernetes.
• Deployments define and manage the number of Pod replicas.
• Services act as an abstraction layer for accessing network Pods.

Kubernetes - Pod Storage (Volumes)

• Volumes provide storage space that survives the lifecycle of individual containers within a Pod.
• Persistent volumes (PVs) are storage resources for Kubernetes.
• Persistent volume claims (PVCs) describe the storage needs of applications.

Kubernetes Pod Resource Usage

• Requests (minimum) and limits (maximum) can control CPU and memory resources of containers within a Pod.
• This guarantees sufficient resources and prevents overconsumption.

Kubernetes Pod Configuration

• ConfigMaps serve as keys for Pod configurations (properties and init values).
• They are similar to .properties or .ini files and support YAML definitions.

Kubernetes Deployments

• Deployments maintain a set of Pods, defining their replicas.
• Kubernetes manages Pod creation and ensures a specified number of running instances.

Other Workload Controller Use Cases

• Deployments: Suitable for deploying stateless applications by specifying a number of replicas to achieve high availability for stateless applications.
• StatefulSets: Manage stateful applications, guaranteeing their persistent state and ordered creation/termination.
• DaemonSets: Configure applications that run on each node, like monitoring/supervision components.
• Jobs: Run tasks with guaranteed completion, like data initialization tasks.
• CronJobs: Run tasks at scheduled intervals for maintenance or other periodic operations.

Kubernetes Services

• Services in Kubernetes act as an abstraction layer, providing external access to Pods.
• Different service types (ClusterIP, NodePort, LoadBalancer, ExternalName) control access within and outside the cluster.

Kubernetes Networking Basics

• Every Pod has a single network interface and a unique IP address.
• Kubernetes uses a cluster network shared among all Pods and services.
• Network Policies allow controlling communication among Pods.
• An ingress network connects pods to the internet and receives external traffic to the cluster.

Kubernetes Receiving External Traffic

• By default, Pods and Services are only accessible within the Kubernetes cluster.
• Admins must configure external connectivity using methods such as exposing individual Pods (not recommended), exposing Services, and using an Ingress resource.

Kubernetes Scaling

• Deployments, StatefulSets, and other Kubernetes resources can be scaled automatically by using HorizontalPodAutoscalers.
• These tools can automatically adjust resource metrics such as CPU and memory, or use custom metrics.