ADVREL01-BP03 Architect for variable demand

Architect to elastically launch resources for variable demand, including the most challenging peak events, like flash crowds or thundering herds.

Implementation guidance

Depending on the advertising channel, such as retail stores, video streaming, or audio apps, loads will peak at different times in different locations. Know your historical load statistics, and adjust load testing scenarios based on historical peaks to determine how the workload performs in unexpected situations and peak demand. With Amazon CloudWatch Real-User Monitoring (RUM), you can collect and view client-side data about your web application performance from actual user sessions in near real-time. CloudWatch Synthetics are configurable scripts that run on a schedule to monitor your endpoints and APIs.

If this a new workload without historical data, load testing is part of this process. Until enough historical data is obtained, use Auto Scaling groups and Elastic Load Balancers (ELB) to meet compute demands and send requests to healthy hosts. Networking demands must also be considered and capacity planned to prevent congestion. For critical workloads, consider private AWS Direct Connect networking to connect to partners or on-premise infrastructure to provide sufficient capacity and more stable latency.

Resources