We deal with the following problem: how to implement a feedback controller on a CPU, with variable computation time, on a CPU that is shared with other tasks so that 1) the task satisfies its Quality of Control (QoC) requirements, 2) the CPU can be efficiently managed to host other concurrent activities. We propose a solution based on the combination of a predictable model of computation and of a scheduling algorithm that enables a fine control on the evolution of the delay. This combination of techniques enables us to set up an optimisation problem where the scheduling parameters are synthesised in order to minimise the CPU utilisation and guarantee the QoC requirements for the control task.