Prototype for a Serverless Business Application

 There are a lot of great tutorials out there on how to build applications based on AWS and in particular on Amplify. But I wanted to try out by myself how much effort it is, to build an application that is a bit closer to a real business desktop app meeting the following requirements:

  • Serverless architecture
  • Based on an RDBMS (relational database)
  • Performing CRUD operations on entities
  • Authentication and authorization
  • Sound exception handling
The result is an application serving currently one entity that can be queried and updated. Here's the architectural diagram:

architecture diagram

The question may be raised, if serverless and using Amplify, why not going NoSQL with DynamoDB? Here are some points that I find important:
  • The majority of software projects is not greenfield. Complex relational data models exist with three digit numbers of tables, numerous (SQL based) use cases and interfaces that can hardly be migrated to a NoSQL data store
  • For business applications in fields like ERP or financials, strict transactional and relational integrity (by foreign keys) is still a key requirement
  • Every day business requests like "What is the summarized monetary value of warehouse items of category A in sales districts South and West" can certainly be answered with a single line SQL select statement. That's not so easy in the NoSQL world.
For this prototype, a PostgreSQL database is in use. That's in AWS terms an Aurora serverless V1. By using Aurora serverless, no running server must be provisioned in the cloud. One downside, this prototype revealed a long cold start time span of the instance once it went to sleep after being idle for 30 minutes. 

The communication of the database with the backend works through the RDS Data API. This simplifies the backend development a lot, compared to going from scratch via JDBC or invoking frameworks like Flask or SpringBoot. There are some important limitations that have to be considered though.

The Lambda functions of the backend are written in Python. This language has earned some reputation in data science. Complex business logic is finally not so different from data science related tasks. With the dictionary type and native JSON support, Python is also a good counterpart for JavaScript based web UIs.

The frontend is implemented in React using the React-Semantic-UI components. 

Finally, for everyone being curious, here's the running application. Use at your own risk. By today (Jan 26 2023) it is up and running. I may have to take it down if it starts costing too much despite the serverless approach. I am planning on publishing most portions of the source code as well.

All comments, hints and improvement requests are highly appreciated.

Comments

Post a Comment

Popular posts from this blog

Build a Secure Database Driven Web Application with Amplify and React: Part 1 - Setup

Image
In my recent post I have published a demo CRUD application utilizing React.js, AWS Amplify and PostgreSQL on Aurora. For everyone who did not want to sign up and take a look for any reason, here's a screenshot: In the next posts, I want to document how easy it is to build a secure serverless web application on top of Amplify using React as a frontend. AWS account AWS command line tools including amplify are installed and connected to the account Installed Node.js (version >=17) and React.js (version >=18.2) An IDE of choice like VS code Initialize the Project To set up the project, have a code editor ready and an terminal window with command prompt. First, we initialize the React application and give it a name like ardemo for Amplify-React Demo.      > npx create-react-app ardemo A directory with the project name will be created, we change to that and can run the intial React application right away:   > cd ardemo > npm start  By default, the node ...
Read more
Image
Time Flies - AWS Aurora serverless has grown up Maybe you remember my demo application leveraging AWS Aurora Serverless v1 on Postgres, using the built in data API and a React frontend. AWS has announced v1 deprecated a while ago and after a postponement, the end of v1 is likely coming by end of next month (March 2025). The AWS community was not so amazed about having to migrate to Aurora v2 as it lacked two important features:  Scaling down to zero which is important for low budget organizations or projects. The data API, providing a highly convenient way to communicate with an RDBMS without any hassle in drivers and connection pooling Now here's the good news: AWS has added both features to Aurora v2 in the meantime, and there is even more: The data API request limit of 1000 requests per second does was removed for v2 . Due to the improved architecture, up- and downscaling performs faster. In particular, scaling up from 0 (the wake up scenario) is faster and results in improved...
Read more