Scaling Real-Time Chat with WebSocket APIs

• A.Use an SQS FIFO queue to store connection IDs and have a Lambda function poll the queue to broadcast.
• ✓ B.Store connection IDs in a DynamoDB table. Use a Lambda function to query all connection IDs and send messages using the API Gateway Management API.
• C.Maintain an in-memory list of connection IDs in a global variable of a single Lambda function.
• D.Use Amazon ElastiCache Redis to store connection IDs and use Redis Pub/Sub for broadcasting.

Why B: The correct answer is B. DynamoDB provides a scalable and cost-effective solution for storing connection IDs because it is a NoSQL database designed for high availability and low latency. The Lambda function can query the entire table to retrieve all connection IDs and then use the API Gateway Management API (via `postToConnection`) to send messages to each client. This approach scales horizontally because multiple Lambda instances can access the same DynamoDB table. Option A is wrong because SQS FIFO queues are not suitable for broadcasting all messages to all connections; they are designed for point-to-point messaging and would require polling, adding latency and cost. Option C is wrong because an in-memory list in a single Lambda instance does not persist across cold starts and cannot be shared across multiple concurrent Lambda instances, leading to data loss and incorrect broadcasts. Option D is wrong because ElastiCache Redis adds operational complexity and cost, and using its Pub/Sub feature would require additional infrastructure; DynamoDB is simpler and more aligned with serverless best practices.

• ✓ A.Store connection IDs in Amazon DynamoDB and use the API Gateway Management API to send messages to all connections.
• B.Use Amazon ElastiCache to cache connection IDs and have Lambda send messages using the Redis pub/sub feature.
• C.Use Amazon SNS to publish messages to all connected clients via the WebSocket API.
• D.Use Amazon SQS to queue messages and have Lambda poll the queue to send messages to all connections.

Why A: Option A is correct because DynamoDB provides a scalable, serverless key-value store to persist WebSocket connection IDs, and the API Gateway Management API allows Lambda to send messages directly to any connected client via its connection ID. This combination efficiently handles the broadcast requirement without managing infrastructure, as Lambda can iterate over stored connection IDs and call the Management API for each message.