微服务应遵循的选项：Spring Boot或AWS Lamba

Question

I come from Java / Java EE background and new to Microservices and AWS; though, I did lot of reading over it. I tried my first microservice with Spring boot and I am impressed with Spring Boot. I also used AWS for past 2 months and I am bit familiar with it now.

Now I am working on a big application and thinking of designing the complete back end (APIs) and would like to use micro services for it - though I would still use only few microservices: One for managing users, and three more with business logic. I will have to also handle authentication.

Based on my research and work I am considering two options:

Option 1: AWS Cognito (for social media login support) + AWS API Gateway + multiple AWS Lambdas (Java Code) + One DynamoDB (preferably otherwise RDS MySQL). My concern is Can I write complex business logic and handle complex tables relations (such as Many-to-Many; eager/lazy fetch, and so on) in AWS lambdas, which are very neatly done in Spring Boot + Spring Data JPA

Option 2: Start with writing business logic in Spring boot, that is, four spring boot applications with either MySQL or DynamoDb as backend; and deploy them on AWS EC2; I can also implement one more Sprint boot application with JWT for authentication My concern here is How to manage authentication (one entry point for all API's) for multiple spring boot applications; If i will face challenge in integrating with AWS API Gateway (for centralized management of Authentication & Authorization); and Will I face challenge in integrating with AWS Cognito if I want social media login

I don't want face lot of complications in later stage, therefore, I need help in deciding.

Many Thanks

Answer 1

When you are having complex business logic/complex relationships which can take more time to execute then you need to be cautious while using combination of Lambda and API Gateway for following reasons:

1. Api Gateway is very useful in case of writing API only if you are sure that your APIs are fast and will bring the response in quick time. Currently, API gateway has the hard limit of timeout after 29 seconds . This means if your API takes more than 29 seconds then it will straight forwardly terminate that request.
2. Something similar goes with lambda as well. Lambdas are specially designed to handle smaller chunks of processing. Currently it supports maximum memory of 3008 MB and timeout of 5 minutes . After 5 minutes lambda will terminate the processing of the request.

So, your Option 1 has these limitations otherwise it's really easy and time+cost effective to use that combo along with AWS Cognito .

Now let's see about your Option 2.

Option 2 is widely used generally for monolithic applications as they are big, more complex and time consuming.

If you are thinking to go with it then I would like to let you know that API Gateway won't be useful because of the timeout hard limit mentioned above. In that case, Authentication can be done using widely available techniques like JWT , OAuth , Basic Auth or even AWS Cognito .

So, in my opinion if you have microservice based architecture then there shouldn't be any issues going with Option 1 provided time performance is honored!

Happy coding!

Answer 2

After investigation I chose AWS lambdas. My application is a event management application and I want event driven architecture. In general question should have been container vs lambdas and I found a good comparison . Starting with Spring is easy because it is so familiar to one with Java background. It is good to write micro services with business logic and exposing as HTTP REST APIs. I was reading domain driven design and it is good to decide micro services size based on boundaries (bounded contexts).

But I also see going with Spring boot I will have to manage authentication (for an event management application we need social media login support), authorization, users, roles, policies, and also finally integrate microservices with API Gateway. AWS Serverless authentication and authorization is very sophisticated . Using NoSQL Database such as DynamoDB follows some of the microservices guidelines such as CQRS. With Lambdas, the boundaries will be too small but in my use case AWS Lambdas suits better.