Testing

We performed the above testing for a unit replica of all files. As another set of testing, we made three replicas of each instead of one and the services performed fine. Just that there was more load on the cluster but the cluster was able to easily handle up to 3 replicas which according to us might be the threshold of the current architecture.

We have use Flask for the development of the API Gateway framework which is easy to use for relatively light workload but ends up breaking if the load is too much. So that is the reason why we have limited the testing to lighter user load but we will improve the architecture to handle a heavy workload. As a part of our testing for every commit, the services was getting started on its own every time and it was working absolutely fine.

In order to check the functionality of the 3 backend microservices by the name of Data Retrieval, Model Execution, and Post-processing, 2 other microservices were built as the purpose of internal testing by the name of the Primary Sender and Final Reciever to check the cross-communication among services and to check the multiple user requests. You can refer the develop branch to see these 2 additional dockerized files built for the sole purpose of testing and not deployment on Kubernetes.