Azure Container Apps Compared With AWS Elastic Container Service

Agenda

• Overview of Azure Container Apps and AWS Elastic Container Services
• High level view of AWS ECS components
• High level view of Azure Container App components
• Comparison of AWS ECS and Azure Container App components
• Features of ECS and Container Apps

Recently Microsoft released their new serverless service for running containers on the Azure platform - Azure Container Apps. It is still in public preview and not ready for production. Microsoft announced the Public Preview of the service in November 2021. This service closely resembles AWS's well matured container deployment service - Elastic Container Service aka ECS.

AWS announced the General Availability (GA) of the ECS services in April 2015. Initially, it started as a container deployment service backed by EC2 instances and supported only Linux containers. Later in December 2016, they added support for Windows Server 2016 containers (beta). In November 2017, Fargate launch type support was added to run containers on a serverless fashion. Fargate is a serverless, pay-as-you-go compute engine for deploying containers that currently supports Elastic Container Service (ECS) and Elastic Kubernetes Service (EKS).

Microsoft launched the public preview of their new container deployment service - Azure Container Apps - in November 2021. It is still in the evaluation phase and not matured enough to run production workloads, you need to wait some more time for the general availability of the service. During public preview, it is only available in limited regions, including Canada Central, West Europe, North Europe, East US, and East US2. You can deploy the Container Apps using Azure Portal, CLI, or ARM template. To deploy with Azure CLI, you need to add an extension for containerapp. (az extension add --name containerapp --upgrade). Since the Container Apps resources are migrating from Microsoft.Web namespace to Microsoft.App namespace, you need to register the namespace using az provider register --namespace Microsoft.App command.

High level view of AWS Elastic Container Service (ECS)

The primary component of the ECS service is the ECS cluster. As per the AWS documentation, a cluster is a logical grouping of tasks or services. It provides the infrastructure support for running containerized applications using Fargate, ECS instances, or the on-premises servers or virtual machines you manage on your data center. Every ECS cluster is deployed in a VPC, which can be a default VPC or a custom VPC created by the user. You can optionally enable the Container Insights feature to collect metrics for many resources, such as CPU, memory, disk, and network. Container Insights also provides diagnostic information, such as container restart failures, that you use to isolate and resolve issues quickly.

Image 1: ECS capacity providers - Fargate, EC2 instances, and external

To deploy your containers on the ECS cluster, you must define the Task Definitions. A Task Definition tells how many containers to run, CPU and memory allocation for containers, port configurations for containers, launch type compatibility, and the OS family. The launch type compatibility can be Fargate, EC2, or external. Task definition also allows configuring the AWS App Mesh based on the envoy proxy that makes it easy to monitor and control microservices. It standardizes microservices communication, and end-to-end visibility and ensures high availability. In a container definition, you can define one or more containers you want to launch together as a single task. While configuring the containers, you can specify the container name, image name, port mappings, health check, and environment variables.

Image 2: Task definition with Fargate launch type

(Reference: https://docs.aws.amazon.com)

Image 3: Task definition with EC2 launch type

(Reference: https://docs.aws.amazon.com) 

A service in ECS allows you to run a specific number of instances of a task definition in the cluster. Service will ensure the high availability of the containers by recreating the containers on failure. Containers will run as a task that is managed by the service. Every EC2 instance runs an ECS container agent that allows containers to connect to the cluster. ECS optimized AMIs comes with the ECS container agent preinstalled.

Image 4: Task placement in EC2

(Reference: https://d2908q01vomqb2.cloudfront.net)

High level view of Azure Container Apps

Azure Container Apps enables you to run your containerized applications in a serverless environment. Behind the scenes, it runs containers on Azure Kubernetes Service with Dapr, Envoy, and KEDA support. It allows you to run your API services, Background processing apps, Event Driven applications, and microservices. It provides Http ingress support, allowing you to access the web applications running in Container Apps using Http or Https. Container Apps can scale containers automatically based on CPU, and memory usage, HTTP traffic, Queues or any KEDA (Kubernetes Event Driven Autoscaling) supported parameters.

Azure Container Apps is still in Public Preview, and when it comes into GA, you may see many changes compared to the current preview version of Container Apps. In Container Apps, you need to create a Container App Environment defining the boundary for individual Container Apps. It logically groups multiple Container Apps that share the same Virtual Network and writes logs into the same Log Analytics workspace. You can deploy Container App Environment in an existing Virtual network also. You may deploy your Container Apps in the same Environment when your applications need to communicate using Dapr and share the same Dapr configuration.

Image 5: Container App Environment

(Reference:https://learn.microsoft.com) 

Application in Container Apps runs in revisions. A revision is an immutable snapshot of the Container App. Creating a container app instance automatically creates a revision for you. Later, whenever you change the container configurations, it creates a new revision for you. You can run revisions in a Single or Multiple mode. In single revision mode, only a single revision of the container app will serve all requests. In multiple mode, you can run multiple revisions of the application and split traffic across those revisions. It is a good strategy for A/B testing and BlueGreen deployments. The total percentage of traffic for all revisions should be 100. Containers in a revision runs inside the Pods. A single pod can run one or more containers inside it.

Image 6: Traffic splitting for revisions

(Reference: https://learn.microsoft.com)

Components of AWS ECS and Azure Container Apps
 

ECS	Container Apps	Description
Cluster	Container App Environment	ECS cluster is a logical grouping of services and tasks. A Container App Environment is a secure boundary for grouping container Apps.
Task Definitions	Revisions	A task definition in ECS defines the task configurations and launch type compatibility. A revision in Container App is an immutable snapshot of the container app that defines the container (Pod) and traffic configurations.
Services	Revision/scale setting	A service in ECS maintains the number of tasks you run in the cluster and defines the load balancer type. You can configure the container scaling by configuring the autoscale setting in the Revisions.
Tasks	Pods	A task is used to run one or more containers inside the cluster. A Container App uses the Kubernetes pods to run one or more containers.
Containers	Containers	A container is a running instance of your application.
ALB/NLB/Classic LB	Http Ingress	ECS can use an ALB, NLB, or classic Load balancer to distribute the traffic across multiple tasks running as part of the Service. Azure Container App uses Http Ingress to accept HTTP requests from outside the VNET.
VPC	VNET	Every ECS cluster runs inside a default or custom VPC. A Container App environment is deployed inside a default VNET, or you can define a custom VNET for the environment.
CloudWatch Container Insights	Log Analytics	CloudWatch Container Insights is a monitoring and troubleshooting solution for containerized applications and microservices. Azure Container App uses a Log Analytics workspace to collect and analyze logs from the Container App Environment.

Feature comparison between AWS ECS and Azure Container Apps

• AWS ECS supports both Windows and Linux container images. But the public preview of Azure Container currently supports only Linux-based container images.
• AWS ECS allows you to register and use external physical/VM instances to run containerized applications. Azure Container apps do not support external instances.
• AWS ECS offers Fargate(Serverless), EC2, and external launch types. But Azure Container Apps currently support only serverless container services backed by AKS.
• AWS ECS uses Task Definitions to configure the container configurations such as image, CPU and memory capacity, launch compatibility, port mappings, health check, environment variables, etc. A revision in Azure Container App defines the container configurations such as image name, CPU and memory, environment variables, and scale settings.
• A task in ECS runs the containers inside the ECS cluster. A Pod in the Azure container app runs one or more containers inside the Container App.
• Azure Container App supports Dapr and Envoy for building and deploying microservices. ECS uses AWS App Mesh backed by Envoy to deploy microservices.
• ECS supports Application Load Balancer, Network Load Balancer, or Classic Load Balancer for applications that require Http inbound support. Azure Container App uses Http Ingress to support Http inbound traffic from outside the cluster.
• ECS clusters runs inside the default VPC, or a user-defined VPC. Azure Container App can be deployed in a default VNET or a custom VNET defined by the user.
• An ECS service allows you to deploy containers either using a rolling update strategy or BlueGreen deployment strategy. Multiple Revision mode and traffic splitting policy in Azure Container App allows you to enable A/B testing and BlueGreen deployment of applications.

Conclusion

AWS ECS is well matured and used by thousands of customers worldwide. AWS added only a few updates on ECS in the last couple of years. Finally, they have added support for external instances using AWS anywhere. As a full-featured container platform service, it can satisfy all kinds of customers who use Windows and Linux based applications. Azure Container App is still growing, and you will see many new features and updates in the coming years. We can expect that Microsoft will add support for running Windows containers on Container Apps soon. AWS ECS and Azure Container App focus on customers who want to build and deploy Microservices on containers backed by Dapr, Envoy, and third-party monitoring services.

Image courtesy

• Amazon Web Services documentation
• Azure Container Apps documentation
• AWS Compute Blog