Serverlose Datenverarbeitung in der Cloud

Einschränkungen und Implikationen für eine IoT-Anwendung am Beispiel von AWS Lambda

Abstract

A popular unique selling point of the cloud is the offer of serverless data processing. With this, customers only pay for those resources they actually use. In addition, resource allocation by the cloud provider is dynamic. From the customer's viewpoint, a serverless cloud service is thus automatically highly available and the risk of under- or over-provisioning is minimized.

The downside of this execution model, however, is that customers leave almost all management expenses to the cloud provider regarding physical and virtual infrastructure. As a result, customers have limited control over the behavior of the services they use, which can translate into concrete disadvantages.

This paper looks at the reasons for these limitations and what identifiable disadvantages they can lead to. The focus here is on the serverless cloud service AWS Lambda, of which the service behavior is examined in more detail in the context of an IoT use case.