chat

try it: https://dwyl.s3.amazonaws.com/index.html

Why?

This repo is designed as a showcase for how to build apps that scale.

We have built chat example apps a couple of times before and the response has been good,
this time our mission is to operate within a very tight set of constraints:

1. No Servers
2. Progressive Enhancement (Works when JavaScript is OFF!)
3. Precisely Predictable (Linear) Performance

What?

Chat. Probably the simplest and easiets to scale implementation you will see ...
unless you work for {{ insert name of silicon valley unicorn messenger app here }}.

How?

Lambda

https://aws.amazon.com/lambda/

• Request Rate Exceeded: http://stackoverflow.com/questions/36826352/aws-lambda-toomanyrequestsexception-rate-exceeded/

S3

We use S3 to render our initial page and host all our static content. https://aws.amazon.com/s3/

• Node.js SDK Examples: http://docs.aws.amazon.com/AWSJavaScriptSDK/guide/node-examples.html
• Working with Folders: http://docs.aws.amazon.com/AmazonS3/latest/UG/FolderOperations.html

API Gateway

API Gateway routes the requests we make from the front-end through to the Lambda function that will process it.

Body Mapping Templates

In order to allow the data submitted by the client to flow through to the Lambda we need to define a "Body Mapping Template".

application/json Body Mapping Template:

{
  ## extract all params in body as JSON;
  "body": $input.json('$'),
  "context" : {
    "method" : "$context.httpMethod",
    "path" : "$context.resourcePath",
    "stage" : "$context.stage",
    "source_ip" : "$context.identity.sourceIp",
    "user_agent" : "$context.identity.userAgent",
    "user_arn" : "$context.identity.userArn",
    "request_id" : "$context.requestId",
    "resource_id" : "$context.resourceId"
  },
  "headers": {
    #foreach($param in $input.params().header.keySet())
    "$param": "$util.escapeJavaScript($input.params().header.get($param))" #if($foreach.hasNext),#end
    #end
  }
}

Once Body Mapping Template is enabled, test using curl:

curl -v -H "Content-Type: application/json" -X POST -d '{"m":"Hello World!","n":"yourname","t":"123456"}' https://r09u5uw11g.execute-api.eu-west-1.amazonaws.com/prod/savemessage

With Authorization Header:

curl -v -H "Content-Type: application/json" -X POST -d '{"m":"1348","n":"yourname","t":"12345678"}' -H "Authorization: eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJpZCI6MSwibmFtZSI6IkFudGhvbnkgVmFsaWQgVXNlciIsImlhdCI6MTQyNTQ3MzUzNX0.KA68l60mjiC8EXaC2odnjFwdIDxE__iDu5RwLdN1F2A" https://r09u5uw11g.execute-api.eu-west-1.amazonaws.com/prod/savemessage

GET messages:

curl https://r09u5uw11g.execute-api.eu-west-1.amazonaws.com/prod/chat

curl -v -H "Content-Type: application/json" -X POST -d '{"m":"Hello World!","n":"yourname","t":"123456"}' https://r09u5uw11g.execute-api.eu-west-1.amazonaws.com/prod/chat

• Overview: https://aws.amazon.com/api-gateway/
• SDK Docs: http://docs.aws.amazon.com/AWSJavaScriptSDK/latest/AWS/APIGateway.html
• HTTP Method: http://stackoverflow.com/questions/35252544/how-to-get-the-http-method-in-aws-lambda
• Passing form data through API Gateway to Lambda: http://stackoverflow.com/questions/32057053/how-to-pass-a-params-from-post-to-aws-lambda-from-amazon-api-gateway which lead to: https://forums.aws.amazon.com/thread.jspa?messageID=673012
• Headers: http://stackoverflow.com/questions/31372167/how-to-access-http-headers-for-request-to-aws-api-gateway-using-lambda
• Velocity Template Language (for mapping): http://velocity.apache.org/engine/devel/vtl-reference-guide.html
• Could not parse request body into json ... https://forums.aws.amazon.com/thread.jspa?threadID=221749

IOT

WebRTC?

Over 50% of browsers (Firefox & Chrome) which means we can cut-out paying for IOT messages for the people who are using good browsers.

http://caniuse.com/#feat=rtcpeerconnection

This also means when we use Electron

How much ($£€) ?

How much does all of this cost...?

Let's break down the cost in the order of the Tech Stack.

S3

https://aws.amazon.com/s3/pricing/

API Gateway

https://aws.amazon.com/api-gateway/pricing/

What is a read/write capacity unit? http://aws.amazon.com/dynamodb/faqs/#What_is_a_readwrite_capacity_unit

Lambda

https://aws.amazon.com/lambda/pricing/

DynamoDB

https://aws.amazon.com/dynamodb/pricing/

Cognito

Amazon Cognito costs $0.15 for each 10,000 sync operations and $0.15 per GB of sync store per month.

https://aws.amazon.com/cognito/pricing/

IOT

$5 per million messages.

A message is a 512-byte block of data processed by AWS IoT – either published to or delivered by the Service. For example, a 900-byte payload is billed as two messages.

https://aws.amazon.com/iot/pricing/

Background Reading

Learning by Doing

• Access HTTP Headers: http://stackoverflow.com/questions/31372167/how-to-access-http-headers-for-request-to-aws-api-gateway-using-lambda
• Cookies on Lambda: http://stackoverflow.com/questions/31851860/access-http-request-headers-query-string-cookies-body-object-in-lambda-with
• Invoke Lambda by HTTP Request: http://stackoverflow.com/questions/29877220/invoke-a-aws-lambda-function-by-a-http-request
• Render HTML in Lambda: http://kennbrodhagen.net/2016/01/31/how-to-return-html-from-aws-api-gateway-lambda/
• Render React on Lambda: https://medium.com/@devknoll/rendering-react-with-amazon-lambda-e4e85a788257

Discussion

We considered using S3 as our primary data store, but soon realized its not that "simple"... see: Why does S3 still not support Appending? https://news.ycombinator.com/item?id=10746969