Igor Dumchykov
Igor Dumchykov

Reputation: 397

The most effective way to implement event scheduling using aws and serverless

Use case: User creates meeting appointment and should be notified in 24 hours/1 hour/5 minutes before appointment. Current implemetation:

  1. When appointment is created, it is saved in DynamoDB with ttl (appointment time - 24 hours)
  2. When ttl is expired, DynamoDB removes this item.
  3. There is a lambda that listens to DynamoDB stream event and is triggered by previous action. There are 3 additional boolean flags in item: 24hours, 1hour, 5minutes. When item is removed before 24 hours, this lambda sets 24hours flag to true (in order to know in the next step that push notification before 24 hours already has been sent) and saves it again with new ttl (appointment time - 1 hour), push notification is sent.
  4. Same as in 2 and 3: ttl expired, lambda sets 1hour flag to true and sets new ttl (appointment time - 5 min) and saves item again, push notification is sent.
  5. Again: ttl expired, push notification is sent.

Concern: DynamoDB does not guarantee that item will be removed exactly when ttl expired.

Is there any another solutions: more efficient than mine?

Current stack: AWS/Serverless Framework/NodeJS.

Upvotes: 5

Views: 515

Answers (2)

Kushan Gunasekera
Kushan Gunasekera

Reputation: 8596

Recently(10-Nov-2022) AWS launched a new service called EventBridge Scheduler. I hope this will fulfill all of your requirements and it's serverless too. What you have to do is create 3 One-time schedule schedules in EventBridge Scheduler. Then set the target as your current AWS Lambda function. Then you don't need Amazon DynamoDB anymore. I hope this will answer your problem.

Upvotes: 2

Hemant Bangar
Hemant Bangar

Reputation: 83

If you just want to use the Serverless stack then this is one of the best ways to do it. But if you want something more functional then you can create a bull queue server [https://www.npmjs.com/package/bull][1] and host it inside an EC2. It provides delayed jobs and queues.

One suggestion that I would like to give is to not directly use lambda to send notifications and instead create a queue and send notifications via that because if your system scales your lambdas will start getting throttled when it has to send thousands of notifications at a single point of time.

Upvotes: 0

Related Questions