The cloud is a distributed architecture providing resources as tiered services. Through the principles of service-orientation and generally provided using virtualisation, the deployment and provisioning of applications can be managed dynamically, resulting in cloud platforms and applications as interdependent adaptive systems. Dynamically adaptive systems require a representation of requirements as dynamically manageable models, enacted through a controller implementing a feedback look based on a control-theoretic framework. We argue that a control-theoretic, model-based architectural framework for the cloud is needed. While some critical aspects such as uncertainty have already been taken into account, what has not been accounted for are challenges resulting from the cloud architecture as a multi-tiered, distributed environment. We identify challenges and define a framework that aims at a better understanding and a roadmap towards control-theoretic, model-based cloud architecture - driven by software architecture concerns.