Deployment Strategy

  • Status: Accepted

  • Last Modified: 2023-07-20

  • Related Issue: 187

  • Deciders: Lucas Brown, Daphne Gold, Sammy Steiner, Billy Daly

  • Tags: ADR

Context and Problem Statement

We need to choose a deployment strategy for the Grants.gov modernization effort that suits our core needs and will host our entire application ecosystem, both the API and frontend layers.

Decision Drivers

  • Reliable: The chosen deployment strategy should be able to offer at least 99.99% monthly application uptime with consistent results.

  • Scalable: We should be able to scale our deployment to meet the demands of our containers while remaining lean.

  • Compatible with IaC: The strategy should be compatible with Terraform as our chosen Infrastructure as Code solution.

  • Ease of use: We prioritize ease of use as well as cost-efficiency. We understand that engineering time spent managing details of infrastructure is a trade off to less prescriptive strategies.

  • Growth oriented: As our applications evolve, we need our chosen deployment strategy to grow with us or be easily swapped.

Options Considered

  • ECS with Fargate or EC2 launch type

  • S3

  • Lambda

Decision Outcome

Chosen option: ECS with Fargate launch type, because it offers the most consistent and easy to use deployment strategy to host both the front-end and API layers of the Grants.gov modernization. Current template infrastructure integrates with ECS and the Fargate launch type.

Pros and Cons of the Options

ECS with Fargate or EC2 launch type

Fargate

Fargate is an AWS serverless compute tool for containers.

  • Pros

    • Run containers without having to manage or provision EC2 instances

    • Removes operational overhead of scaling, patching, securing and managing servers

    • Integrates with AWS Cloudwatch or other third party metrics tools

    • Secure, running in dedicated runtime environments

    • Scalable means pay for usage, not for reserved or wasted space

    • Fargate Spot instances further reduce costs and are viable for our use-case

  • Cons

EC2

Secure and resizable compute capacity for virtually any workload.

  • Pros

    • Ability to granularly manage and provision resources of environment

    • Can connect to an EC2 instance easily via SSH if necessary

    • Many instance types to choose from to meet our needs

    • Reliable, scalable and on-demand

    • Compatible with many other tools in the AWS arsenal

  • Cons

    • Requires detailed management and provisioning of environment

    • Mismanagement of environment can greatly increased costs

    • Less friendly for engineering teams that are not DevOps dedicated

A note on ECS: It is possible to run EC2 instances to host our Docker containers without using ECS orchestration. However, since ECS is a free service provided by AWS and we would only pay for the underlying resources, forgoing ECS and an orchestration tool isn't an appealing strategy.

S3

Object storage built to retrieve any amount of data from anywhere.

  • Pros

    • Suitable to host static websites, our current use-case

    • Highly scalable with unlimited storage space

    • Extremely cost-effective with pay-as-needed pricing model

    • Highly durable with storage redundancy in multiple locations

    • Easy to use interface with static website hosting options

  • Cons

    • Only suitable to host static websites, making it a difficult choice for a rapidly growing front-end in development

    • Limited customization options

Lambda

Run code without thinking about servers or clusters

  • Pros

    • Auto-scaling has limitations, and pay-per-request methodology is ultra lean

    • No need for redundancy in multiple Availability Zones

    • Run code without provisioning or managing any infrastructure

    • Scalable to meet high demand

  • Cons

    • Ineffective for long-running processes, maximum duration of 15 minutes

    • Deployment package maximum size is 250 MB

    • Max Docker image size is 10 GB

    • Deploying serverless applications might require project restructuring or additional tools and ramp up

    • Less performant than alternatives with lack of dedicated resources

Interesting read on building a static React front-end hosted in S3 with Lambda back-end: Server-side rendering for React in AWS Lambda

Last updated