Deploying API in AWS using Lambda

[Rumman]

*** *** *** *** *** *** *** *** *** *** ***

In the question Topic-Based – Lambda (SA-Associate)

Category: CSAA – Design High-Performing Architectures

A game development company operates several virtual reality (VR) and augmented reality (AR) games which use various RESTful web APIs hosted on their on-premises data center. Due to the unprecedented growth of their company, they decided to migrate their system to AWS Cloud to scale out their resources as well to minimize costs.

Which of the following should you recommend as the most cost-effective and scalable solution to meet the above requirement?

—- —- —-

The options are given

– Set up a micro-service architecture with ECS, ECR, and Fargate.

– Host the APIs in a static S3 web hosting bucket behind a CloudFront web distribution.

– Use a Spot Fleet of Amazon EC2 instances, each with an Elastic Fabric Adapter (EFA) for more consistent latency and higher network throughput. Set up an Application Load Balancer to distribute traffic to the instances.

– Use AWS Lambda and Amazon API Gateway.

*** *** *** *** *** *** *** *** *** *** ***

My understanding is along with using AWS Lambda and Amazon API Gateway; this can also be done using Set up a micro-service architecture with ECS, ECR, and Fargate, as a docker based web API application. Moreover, this can also be deployed like an on-demand task using Fargate Task; considering Lambda’s cold start time for a traffic-heavy website, it might be even better to use EC2 and ECS based solution. So why this is not a valid answer?

• This discussion was modified 3 years, 5 months ago by  Rumman.

[Carlo-TutorialsDojo]

Hello muhammad-r,

Thanks for your feedback.

Yes, you can dockerize a RESTful web service using ECS or even host it on EC2. However, the main concern for the scenario is cost-efficiency and scalability. Using ECS and EC2 requires capacity planning ahead of time to support the load. Among the given choices, API Gateway and AWS Lambda is the most suitable answer.

Lambda’s cold start time isn’t much of a concern for an application with high traffic because subsequent requests can re-use the container where previous requests have run. It takes about 30-45 minutes before a function goes into “sleep mode” after execution. It means that within that duration, the probability where a container is invoked again before it can even begin to shut down is high.

Let me know if that answers your question.

Regards,

Carlo

• This reply was modified 3 years, 5 months ago by  Carlo-TutorialsDojo.