Cost efficiency under mixed serverless and serverful deployments

Function as a Service (FaaS) is an integral part of the serverless computing paradigm. It offers a true pay-per-use billing model and releases developers from the burden of managing the application stack. A discussion on whether and when this model is more appropriate for cloud computing users in terms of accruing costs compared to the more "traditional"delivery models has already been started by existing works. However, by treating this subject as a regular service selection problem, these approaches fail to exploit the space created by distributing the load between simultaneous FaaS and non-FaaS deployments of an application in a hybrid deployment model. This work aims to provide the means for application owners to decide which deployment scenario is cost optimal for their needs. In case this scenario is a hybrid deployment, the proposed approach also determines the optimal number of virtual machines that will need to be provisioned. An extensible and configurable FaasSimulator open source tool is presented for these purposes.