Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/ti-infinite/GSMInfrastructure/llms.txt

Use this file to discover all available pages before exploring further.

GSM Infrastructure ships with a layered cost-control strategy that combines an AWS Budget ceiling, real-time SNS email alerts, and an automated EC2/ECS start-stop scheduler. Together these controls provide visibility into spend before it exceeds your target and eliminate idle instance-hours during off-hours — all configured declaratively through CloudFormation parameters so each environment can carry its own limit.

​
AWS Budget

The AwsCostBudget CloudFormation resource provisions a monthly COST budget automatically named: 
{env}-{appName}-monthly-budget
The budget tracks actual (not forecasted) spend and fires a notification when your account crosses 100% of the BudgetLimitUSD parameter value. The default limit is $30 USD, but you should override this per environment in your GitHub Actions variables (BUDGET_LIMIT). 
# infrastructure/template.yml — budget resource (excerpt)
AwsCostBudget:
  Type: AWS::Budgets::Budget
  Properties:
    Budget:
      BudgetName: !Sub '${Environment}-${AppName}-monthly-budget'
      BudgetLimit:
        Amount: !Ref BudgetLimitUSD
        Unit: USD
      BudgetType: COST
      TimeUnit: MONTHLY
    NotificationsWithSubscribers:
      - Notification:
          NotificationType: ACTUAL
          ComparisonOperator: GREATER_THAN
          Threshold: 100
          ThresholdType: PERCENTAGE
        Subscribers:
          - SubscriptionType: SNS
            Address: !Ref NotificationSNSTopic
AWS Budgets evaluates spend up to three times per day. The alert may not fire the instant you cross the threshold — expect a delay of several hours at most.

​
SNS Notification Topic

The infrastructure stack creates an SNS topic named: 
{env}-{appName}-notification-alerts
An email subscription is added to the topic at deploy time using the AlertEmail CloudFormation parameter (set via the ALERT_EMAIL GitHub variable). AWS sends a subscription confirmation email to that address immediately after the stack is created or updated.
The budget notification will not be delivered until the recipient clicks Confirm subscription in the confirmation email. Check your inbox (including spam) right after the first deploy.
The SNS topic also carries an IAM topic policy that allows events.amazonaws.com to publish to it, keeping the door open for future EventBridge-to-SNS integrations.

​
Automated Scheduler

The start/stop scheduler is the single largest lever for cost reduction. A pair of EventBridge Schedules invoke Lambda functions on a weekday cron to:
  1. Stop — scale all four ECS services to desiredCount=0, disassociate the Elastic IP, then stop the EC2 instance.
  2. Start — start the EC2 instance, wait until it is running, reassociate the Elastic IP, then scale ECS services back to desiredCount=1.
Stopping the EC2 instance nightly eliminates EC2 instance-hour charges for the entire off-hours window. Disassociating the EIP before shutdown avoids the AWS charge for an idle (unassociated) Elastic IP, which applies whenever an EIP is not attached to a running instance. See Scheduler for the full schedule configuration, override procedures, and retry policy details.

​
Cost Optimization Tips

Set TaskNumberDesired=0 (GitHub variable TASK_NUMBER_DESIRED) during initial setup and image build phases. ECS will not attempt to place tasks — and you will not be charged for container runtime — until you are ready to push your first images to ECR.

Use t4g.medium (arm64)

The default Ec2InstanceType is t4g.medium. Graviton-based instances offer roughly 20% better price-performance than equivalent x86 instance types. Both Lambda functions in the scheduler stack are also deployed on arm64 to stay consistent.

CloudFront PriceClass_100

The CloudFront distribution uses PriceClass_100, restricting edge locations to North America and Europe. This is the lowest-cost price class. If your users are in other regions, evaluate whether the latency trade-off justifies upgrading to PriceClass_200 or PriceClass_All.

Short CloudWatch log retention

Backend ECS logs (/ecs/{env}-{appName}-backend) are retained for 7 days. Scheduler Lambda logs are retained for 14 days. CloudWatch Logs storage is billed per GB — keeping retention windows short prevents unbounded accumulation.

Per-environment budget limits

Set a conservative BudgetLimitUSD in development and QA (e.g., $15) and a higher value in production. This way, unexpected spend in lower environments triggers an alert before it silently balloons.
If you are running a purely development environment that is only used during business hours, combine TaskNumberDesired=0 with an aggressive stop schedule (e.g., stop at 6 PM COT) to maximize savings. The EC2 instance is the dominant cost driver — every hour it is stopped saves you instance-hour charges.

Build docs developers (and LLMs) love

Get started for freeTalk to us