Operating applications in the cloud requires visibility into both infrastructure performance and user activity. AWS addresses these needs through a set of monitoring and messaging services. Amazon CloudWatch is the central monitoring service, collecting metrics, logs, and events from AWS resources and applications. CloudWatch provides dashboards for visualization, alarms that trigger when metrics cross thresholds, and automated actions that can respond to those alarms, such as invoking an Auto Scaling policy or invoking a Lambda function. With CloudWatch, operators gain a unified view of system health and can detect and respond to issues in near real time.
While CloudWatch focuses on performance and operational telemetry, AWS CloudTrail focuses on auditing and governance. CloudTrail records every API call made in an AWS account, capturing who made the request, when it was made, which service was called, and which resources were affected. These event histories are stored in log files and can be used for compliance reporting, security analysis, change tracking, and troubleshooting. Together, CloudWatch and CloudTrail offer complementary perspectives: CloudWatch shows how systems behave, while CloudTrail shows who did what and when.
Many modern applications are built as collections of decoupled services that communicate through messaging. Amazon Simple Notification Service, or SNS, implements a publish-subscribe model in which publishers send messages to a topic and multiple subscribers receive them. SNS is well suited to fan-out scenarios such as sending notifications via email, SMS, or HTTP endpoints. Amazon Simple Queue Service, or SQS, provides message queuing for distributed systems and microservices. Producers place messages into a queue, and consumers retrieve them asynchronously, which helps smooth out bursts in demand and ensures that work is not lost even if a consumer temporarily fails. Combining SNS and SQS lets teams design reliable, event-driven architectures that scale independently and recover gracefully from component failures.