Node-level response time feedback loops to ease QoS control in “as a Service” architectures

Many ICT (Information and Communications Technology) functionalities are nowadays offered “as a Service” (aaS), i.e., by collecting user requests and routing them through a dynamically chosen set of network nodes. In such systems, enforcing the desired Quality of Service (QoS) is particularly complex. We argue that a prominent reason for this problem is that aaS frameworks invariantly miss a simple but crucial component, namely a layer of node-level response time control loops. We propose a solution to fill this gap by exploiting periodic event-based feedback linearisation and PI control. We demonstrate the potential of the said solution on a real application and briefly discuss the numerous implications to address in subsequent works.