PayU is a payment solution provider, and this solution was particularly built for Marketplaces that have different infrastructure needs at different periods of time, e.g. during holidays or festive seasons, there will be a lot of activity and transactions. Hence the solution needed to scale up significantly during the peak season and downscale during the lean period. After executing some of the initial poc we narrowed down on Serverless Computing as the preferred choice. Serverless computing allows us to build and run applications and services without thinking about servers keeping us focused on functionality development. With this approach, we worked on developing individual AWS Lambdas, and to make the solution more cost-effective, we combined various Lambdas in AWS Step Functions, building a simple state machine to orchestrate the flow. Step Functions coordinate the components of distributed applications and microservices at scale, using visual workflows that are easy to build.

The Figure below illustrates one of the flows that was designed for this case study, the various lambdas are combined using the Step Functions and these state machines were triggered using another trigger function lambdas which were called whenever there is a purchase request. We made sure that Lambdas performed one and only one function at a time to increase their reusability. DynamoDB was used for storing the data whenever required, and for 3rd Party processing we used webhooks to get the status updates in an async way.

We also used DynamoDB Streams to create event-driven flows that were required for analytics and other sidecar operations like Notification Service and storing the operational history related to purchases.