Choosing a modern application strategy

Organizations standardize on automation technologies to take full benefit of the cloud. The strategy and tools used by infrastructure, platform, and DevOps teams often drives decisions.

For example, organizations have choices for tooling that automates the creation, configuration, and maintenance of infrastructure, resources, and workloads. These tools include infrastructure-as-code (IaC) tools such as Terraform, CDK, CloudFormation, or other community tools. 

Organizations that use Kubernetes might use Kubernetes-based tools for automation using GitOps tools such as ArgoCD and Flux and Kubernetes API-based cloud provisioning tools like AWS Controllers for Kubernetes (ACK) or the CNCF Crossplane project.

These tooling choices extend to tools that extend to security, testing, networking, observability, performance, and more. For each of these categories, there is a stack of available tools.

These automation tools often have built in integrations and accelerators for compute choices. For example, AWS Serverless Application Model (SAM) is optimized for serverless developers to build and quickly deploy Lambda functions. Other tools, like ArgoCD, automate the deployment and configuration of Kubernetes workloads.

As customers increase workloads, it can be a burden to support a large number of tooling choices. We recommend standardizing a set that support the most workload patterns.

It is important to evaluate the most appropriate compute option on a workload-by-workload basis within the default strategy. You should always strive to achieve the desired performance, security, and cost benefit for each workload.

A good standardization strategy will allow for different use cases such as microservices, modernized monoliths, event driven architectures, tools built by operation teams, and data processing workloads, such as machine learning, batch processing, and stream processing.

These workloads have different architectural characteristics. The strategy adopted should allow flexibility to support all the stages of a developer or data scientist workflow.

• Application Developer: Needs to run multiple environments such as development, prototyping, test, staging, and production.
• Data engineer: These workflows involve streaming and acting on large data models, cleaning, training, running inference with models, and building applications and data pipelines that experiment with the data, including Jupyter Notebooks.
• AI/ML scientist: Generative Artificial Intelligence (AI) Large Language Models (LLM) may require specialized compute instances such as AWS Trainium or other GPU-based architecture.
  

Applications do not exist in isolation. They are supported by technologies such as databases, messaging, streaming, orchestration, and other services. An effective modern app development strategy requires integration with these services. Managed integrations simplify operational overhead as much as management of the underlying infrastructure.

AWS serverless compute options, such as AWS Lambda, are integrated into the AWS ecosystem. AWS Lambda can subscribe to events from more than 250 other services.

AWS-managed offerings for Kubernetes also provide integrations with many AWS managed offerings. For example, you can use AWS Controllers for Kubernetes to provision native AWS resources using a Kubernetes API.  In addition, Kubernetes has a rich ecosystem, offering integration with numerous open-source projects. Amazon EKS works with many other AWS services to provide additional solutions for business challenges.

Many organizations need to create experiments to validate ideas. The ability to provide an environment where you can quickly write, deploy, and validate ideas is essential for a healthy environment.

This environment is often overlooked when developing a modern app development strategy, but the ability to innovate may depend on it. Enabling teams to use services that allow builders to rapidly build, test, and iterate help in discovering new business opportunities and receive feedback faster.

Serverless compute options like AWS Lambda and AWS App Runner are optimized to enable organizations write code quickly, deploy it, and change it. These capabilities provide useful options for doing fast prototyping work that doesn’t require making a lot of choices upfront. Developers can quickly turn an idea into a modern, working application. The AWS serverless compute option provides choices to prototype with minimal cost or operational overhead.

Other options can be used for prototyping: for ECS, you can create dedicated clusters and use Fargate or dedicated nodes for prototyping, for Kubernetes, platform teams can create dedicated clusters or have namespaces within a cluster dedicated to prototyping teams. Open source projects such as CNCF BuildPacks or Knative can be used to simplify the configuration experience.