AWS Well-Architected pillars
This solution uses the best practices from the AWS Well-Architected Framework
This section describes how the design principles and best practices of the Well-Architected Framework benefit this solution.
The machine-learning lifecycle is the iterative process, with instructions and best practices, to use across defined phases while developing an ML workload. It adds clarity and structure for making a machine learning project successful. The Well-Architected machine learning lifecycle superimposes the Well-Architected Framework pillars to each of the machine learning lifecycle phases illustrated in the center of the following figure.
The Well-Architected machine learning lifecycle
Operational Excellence
This section describes how we architected this solution using the principles and best practices of the operational excellence pillar.
The QnABot on AWS solution pushes metrics to Amazon CloudWatch at various stages to provide observability into the infrastructure; Lambda functions, AI services, Amazon S3 buckets, and the rest of the solution components. Continuous integration and continuous delivery (CI/CD) and infrastructure deployment are managed in code through AWS Amplify. Data processing errors are added to the Amazon Simple Queue Service (Amazon SQS) queue and displayed in the application layer for user response.
Security
This section describes how we architected this solution using the principles and best practices of the security pillar.
-
Content designer UI app users and the Amazon Lex client are authenticated and authorized with Amazon Cognito.
-
User permissions to app accounts are managed in the Amazon DynamoDB.
-
All inter-service communications use AWS Identity and Access Management (IAM) roles.
-
All multi-account communications use IAM roles.
-
All roles used by the solution follows least-privilege access. That is, it only contains minimum permissions required so the service can function properly.
-
Communication end user and Amazon API Gateway uses Bearer token generated and handed by Amazon Cognito.
-
All data storage including Amazon S3 buckets have encryption at rest.
Reliability
This section describes how we architected this solution using the principles and best practices of the reliability pillar.
-
The solution uses AWS Serverless Services wherever possible (examples Lambda, API Gateway, Amazon S3, and Amazon Lex) to ensure high availability and recovery from service failure.
-
The solution protects against state machine definition errors by having automated tests performed on the solution.
-
Data processing uses AWS Lambda functions. Data is stored in DynamoDB and Amazon S3, so it persists in multiple Availability Zones by default.
Performance Efficiency
This section describes how we architected this solution using the principles and best practices of the performance efficiency pillar.
-
The solution as mentioned earlier uses serverless architecture throughout this solution.
-
The solution can be launched in any Region that supports AWS services in this solution such as: AWS Lambda, Amazon API Gateway, AWS S3, Amazon Lex, Amazon Kendra, and Amazon Comprehend.
-
The solution is automatically tested and deployed every day. As well as reviewed by solutions architects and subject matter experts for areas to experiment and improve.
-
The QnABot on AWS CLI supports the capability to import and export questions and answers from your QnABot setup are designed to reduce IT overhead for maintenance and upkeep.
Cost Optimization
This section describes how we architected this solution using the principles and best practices of the cost optimization.
-
The solution uses serverless architecture therefore, customers only get charged for what they use.
-
The compute layer defaults to AWS Lambda, so it provides pay per use. DynamoDB indexes are selected to reduce throughput cost for queries.
-
The solution provides an option to the user to use more advanced AI/ML services. Services such as Amazon Kendra and Amazon SageMaker are optional and can be turned on or off to reduce the cost for users who don’t intend to use these features.
Sustainability
This section describes how we architected this solution using the principles and best practices of the sustainability pillar.
-
The solution utilizes managed and serverless services, to minimize the environmental impact of the backend services. A critical component for sustainability provided by the solution is maximizing the usage of the AWS AI services. The solution Serverless design (using Lambda and DynamoDB) and the use of managed services (such as AWS Amplify) are aimed at reducing carbon footprint compared to the footprint of continually operating on-premises servers.
