How to Design a Serverless Application - Survey App Solution Architecture

[@sakiboutchagbele8717]

I think the video is very good. I really like this new format with an end-2-end approach, from requirements, application architecture to implementation with AWS and the Serverless Framework. Please keep it that way, great!

[@CompleteCoding]

Great to hear! I'll probably be doing similar videos like this in the future. I might find and existing software tool and work out how I would make it Serverlessly.

[@PeteWhelan]

Congratulations Sam - You hit 5000 subscribers!

[@CompleteCoding]

Thanks Pete, I'm so happy to make it this far and super excited to see how far it can go.

[@naveensaini1618]

Hey Sam, very very thanks for this video, I am eagerly waiting for this video after so long, I hope 2021 will give you bounce back startup towards your journey

[@CompleteCoding]

Thanks so much. I hope your 2021 is awesome too

[@PeteWhelan]

Thanks Sam - This has helped me with documentation on my projects.

[@CompleteCoding]

Glad it helped!

[@tn9753]

Hi Sam. I enjoy & appreciate your videos. I noticed a mention in the comments about a Total Cost of Ownership video. I second this request. Keep up the good work.

[@CompleteCoding]

I've had a few people ask for this now so it'll definitely be coming soon

[@RahulAhire]

5:35 what you think is the best way to use Graphql in serverless - Using AppSync or our own custom Graphql lambda function with proxy integration

[@CompleteCoding]

I think that appsync is the best way to get this set up. It's easy and you know that everything will be set up properly

[@RahulAhire]

@@CompleteCoding don't you think it's too much costly as compared if we used proxy integration?

[@CompleteCoding]

​@@RahulAhire AWS and Serverless costs are always an interesting discussion. Yes, the AWS costs to use Appsync will be higher than using a lambda proxy, but those aren't the only costs. Developers are often far more expensive than AWS Costs. Even with 10,000 users, if they each did 100 queries/updates a month that is 1M requests which would be about $21 in AppSync. With Lambda proxy it might be $2. If your developer costs $50/hr and they spend one day creating the custom lambda proxy then it will take 20 months to make it worth while. You also have added costs whenever you need to update the graphql schemas and new developers take longer to understand what's going on. I'm thinking of doing a "true cost of software ownership" video soon

[@RahulAhire]

@@CompleteCoding this is really interesting and honestly I forgot this to consider. I'd love to see a video on it from you. Between thanks for explaining 👌🏻

[@luismenendez2040]

Very grateful with your content !!

[@CompleteCoding]

I appreciate that! It's great hearing that my content is helping people

[@bluestar4207]

I have questions. What is the purpose of the cloudfront and s3

[@CompleteCoding]

S3 is the bucket that stores the react app. Cloudfront does a lot of things but basically caches the react app locally to users so page loads are faster.

[@peterkulik5943]

Why do you plan to use two different technologies (aurora and dynamodb) to store data in one solution? Why don't you use only aurora?

[@CompleteCoding]

Personally I've got a lot more experience with dynamo and the tips and tricks to get it working optimally. I'm getting to try out aurora v2 and if it works as claimed then it could become my go to storage, but for now that's dynamo.

[@jasper5016]

This is looking great.

[@CompleteCoding]

Thanks

[@metalheadmask]

Great work. Now, what if we change the design statement. Let's say, the client would like to be discrete with the form input data, how do you see this architecture holding up?

[@CompleteCoding]

Are you talking about switching out the google forms survey for a different mechanism for receiving and storing user responses which would be more secure? If so, you would need two new tables, one for the actual questions/options that make up the survey. The second would be a database to store the responses. Both of these could be accessed through the same GraphQL services. You could restrict the access to read response data to admins or the person who submitted the survey. Not sure if this answers your question?

[@metalheadmask]

@@CompleteCoding Cool, you got it right. I was concerned of the data from response forms built by Google. But yeah there are further questions on deploying machine learning. I'll try to catch up if you have a web address.

[@CompleteCoding]

@@metalheadmask My email is in my "about" section

[@vladanulardzic5859]

What is the purpose of Aurora Users and what it the purpose of User Pool ?

[@CompleteCoding]

Aurora is used as SQL has much more flexible querying. With Dynamo you can only query on two fields. (where partition key = X and sort key (=, starts with, greater than) Y. You can use compound keys to make it a bit better but it is still limited. With SQL you can query more efficiently on any of the columns. Since an admin could query by country, age, job and hobbies - all at the same time, SQL is a better fit. You could go with some other tool (ElasticSearch) but that will involve getting servers involved which is what we're trying to avoid.

[@james.sweetland]

Very helpful Sam!

[@CompleteCoding]

Glad it was helpful!

[@prerakhere]

Not sure if real apps are made without the vpc

[@darshitgajjar5199]

great sir, make more videos

[@CompleteCoding]

I'm definitely planning to