Lessons from AWS Cloud Powering Amazon.com’s ‘Biggest Day Ever’ Amazon.com is just one customer of Amazon Web Services Inc. (AWS), which powered the online store during the record-setting Prime Day 2016. Jeff Barr, an AWS spokesperson, reported today on how the cloud platform reacted to the increased demand for global ecommerce due to the sale. This sale reportedly accounted approximately three-quarters (or 3) of all U.S. consumer online ecommerce as of July 11. Amazon.com is a customer of AWS. However, the two entities communicate regularly with each other, which allowed AWS to allocate more EC2 computing resource to handle store Web traffic that reached new heights, generating 85 million clickstream log entries. The sale of EC2 computing resources prompted an increase in 38 services. These services range from analytics (which saw a 1,661 per cent increase in events) and storage and content delivery. Barr stated in a blog post that AWS allows customers to increase the power required for big events like Prime Day. It also makes it possible to acquire this capacity in a more flexible, cost-effective way. AWS handles all of the heavy lifting necessary to create an event online at this scale. This allows Amazon’s retail team to focus on providing the best possible customer experience. Barr did not address many media reports about technical glitches that were reported during the sale, including the inability to add products into customers’ shopping carts as reported by Money. A spokesperson for Amazon stated that an issue with Lightning Deals checkouts was fixed during the sale, according to a US News & World Report article. Barr stated that while they did not comment on the reported problems, the Amazon retail team was pleased that Prime Day was over and was ready to rest, but they shared some lessons with me. These lessons, which could be useful to other customers who are planning big events, include:
Preparation and testing are crucial. You can use historical metrics to forecast and model future traffic and to estimate your resources needs accordingly. GameDay exercises prepare you for failures — deliberately breaking different parts of the infrastructure to simulate various failure scenarios. (Read Resilience Engineering — Learning To Embrace Failure to find out more about GameDay exercises at Amazon).
Automate: Automate all your tasks and reduce manual effort. You can take advantage of services that scale automatically according to demand. Route53 will automatically scale your DNS, Auto Scaling will scale your EC2 capacity based on demand, and Elastic Load Balancer will automatically failover and balance traffic across multiple regions or availability zones (AZs).
Monitor: Make use of Amazon CloudWatch alarms and metrics. CloudWatch monitoring allows you to keep track of your usage and ensure that your customers have the best experience possible.
Think Big: AWS provided the team with the resources necessary to create another holiday season. Your ability to scale large events is dependent on your infrastructure.
Barr advised AWS customers to contact AWS to discuss communication and AWS support plans in preparation for any large-scale, temporary events.