+1 on this. The paper doesn't really address the most typical usecases of Serverless which are typically API backends, streaming data processing (feeding off Ki

+1 on this. The paper doesn't really address the most typical usecases of Serverless which are typically API backends, streaming data processing (feeding off Kinesis/Kafka/similar), and reactions to one off-ish events like object uploads to storage or events from other systems. While folks like those building Pywren have found success with highly parallel workloads, FaaS still isn't the typical way people go after distributed computing workloads, and no one from the vendor side is pushing this. - munns - Lead Dev Advocate - AWS Serverless