“Amazon Kinesis can collect and process hundreds of terabytes of data per hour” –> A little fun with the pricing tool reveals you’ll also be paying $millions per hour. To put that in perspective, just one such customer would represent more than 1% of Amazon’s total revenue.
We’re using it to stream metrics off our clusters. Seems good so far. Kafka-as-a-service with a custom api. No major hiccups yet.