Guidance for Demand Forecasting & Planning on AWS
Overview
How it works
This architecture shows how data is collected from your databases and ecommerce sites, processed, and exported to a dashboard for visibility and forecasting.
Well-Architected Pillars
The architecture diagram above is an example of a Solution created with Well-Architected best practices in mind. To be fully Well-Architected, you should follow as many Well-Architected best practices as possible.
Operational Excellence
This Guidance is designed to provide you with the information necessary to help you understand your internal business state. For instance, each component sends logs and metrics to Amazon CloudWatch for monitoring. And DynamoDB is used to process information, such as last forecast completion time and number of items analyzed, providing transparency to your business users. You can also choose to deploy this Guidance with AWS CloudFormation that allows for small and frequent changes, and adapt it to fit within a continuous integration and continuous delivery (CI/CD) pipeline.
Security
This Guidance provides a selection of capabilities that helps ensure you have robust identity management in place. With Amazon S3, all buckets have encryption enabled and are configured to restrict access for only those services that should interact with it. The other services in this Guidance use AWS Identity and Access Management (IAM) policies with least-privilege access, allowing users to connect and complete only the necessary actions. QuickSight is provisioned with a login page for business users, and if you deploy the option of a webpage for forecasts with DynamoDB, you can use Amazon Cognito for authentication.
Reliability
This Guidance supports a reliable architecture for each application level. The components that process data, such as AWS Lambda, AWS Glue, Step Functions, and Athena are serverless, reducing concerns with scalability and scaling. In the data layer, this Guidance uses Amazon S3 that provides 11 9s of durability and DynamoDB that scales automatically to adapt to the application's load. And Forecast and QuickSight are fully managed services that provide automated recovery from failures and scalability.
Performance Efficiency
The services selected for were purpose-built for this Guidance. The fully managed services, such as QuickSight, will adapt its capacity for the number of interactions, providing performance as it scales. DynamoDB auto-scales horizontally, allowing consistency in performance even with peak loads. You can experiment with this Guidance by adjusting Lambda and AWS Glue to process data faster and according to the needs of retraining the model. SageMaker Canvas will explore and process a series of adjustments based on the user’s data to achieve the best performance without the need for the user to understand machine learning extensively.
Cost Optimization
The components in this Guidance are serverless, providing a pay-as-you-go approach, avoiding oversized, provisioned resources to help you keep the costs related to the number of completions and the amount of data to be processed. From the data perspective, ETL is processed by a scheduled AWS Glue job, stored in Amazon S3, and read by Athena, avoiding spends with servers. SageMaker Canvas will be retrained only when scheduled, avoiding costs with this process as well. QuickSight provides cost per active user, allowing you to start with costs only for dedicated analysts.
Sustainability
This Guidance uses a serverless first approach, which means that there are no compute idle resources. The fully managed services in this Guidance scale as demand grows, reducing and optimizing the number of resources running. For example, AWS Glue will process on schedule, turn on its resources, process ETL, store it on Amazon S3, and turn the servers and the resources down. This directly reduces energy consumption and the impact of your carbon footprint.
Disclaimer
Did you find what you were looking for today?
Let us know so we can improve the quality of the content on our pages