Abstract-New dynamic cloud pricing options are emerging with cloud providers offering resources as a wide range of CPU frequencies and matching prices that can be switched at runtime. On the other hand, cloud providers are facing the problem of growing operational energy costs. This raises a trade-off problem between energy savings and revenue loss when performing actions such as CPU frequency scaling. Although existing cloud controllers for managing cloud resources deploy frequency scaling, they only consider fixed virtual machine (VM) pricing. In this paper we propose a performance-based pricing model adapted for VMs with different CPU-boundedness properties. We present a cloud controller that scales CPU frequencies to achieve energy cost savings that exceed service revenue losses. We evaluate the approach in a simulation based on real VM workload, electricity price and temperature traces, estimating energy cost savings up to 32% in certain scenarios. I. INTRODUCTIONWith the wide range of VM types, heterogeneous infrastructure and different computing environments including VMs and containers [1], estimating the performance of provisioned resources is becoming increasingly challenging [2]. New cloud pricing schemes are emerging where resources are priced based on the delivered performance. For example, the CPU frequency provided to a VM at runtime determines the price [3], with high CPU frequencies being more expensive. We call this model performance-based pricing and it is used in production by cloud providers, such as ElasticHosts [4]. Though this approach mainly targets users, it could also be used by cloud providers to control their energy consumption. Energy consumption of data centers is becoming a major issue, accounting for 1.5% of global electricity usage [5]. Furthermore, modern clouds may consist of geographically-distributed data centers influenced by dynamic local factors, such as real-time electricity prices [6] and temperature-dependent cooling [7], that we call geotemporal inputs.We call the subsystem of the cloud, that determines the actions to allocate and manage the VMs, a cloud controller. Adapting the cloud controller to geotemporal inputs through actions, such as CPU frequency scaling, raises a trade-off problem between the potential energy savings and service revenue losses incurred under performance-based pricing. This is the challenge at the core of this paper.Frequency scaling is a power management technique commonly used to lower the operating frequency of hardware resources in order to reduce power consumption [8]. However, frequency reduction may degrade the performance of resources. Depending on the workload characteristics, workload