In Chap. 4 , effective mitigation approaches for performance variability based on concepts of system and adaptive scenarios have been presented. However, the goal of full timing guarantees has not yet been achieved in these approaches. Therefore, here we further propose a novel workload-dependent multi-timescale mitigation approach to ensure timing guarantees for highly dynamic workloads, tackling the variability from the architecture and software layers in sub-millisecond time granularity. As motivated in Chap. 3 , the specific problem formulation of mitigating performance variability with full timing guarantees has not been achieved up to now in state-of-the-art. Nevertheless, our approach will show that full timing guarantees can be achieved by workload-dependent approaches. In the following two chapters, we present the approach in full detail. In this chapter, the principles and design of the experimental approach are elaborated, while in Chap. 6 the simulation we performed in testing this approach is discussed. This chapter is structured as follows: Sect. 5.1 describes the motivation and design principles of this approach. The rest of this chapter introduces the salient features of this approach. In Sect. 5.2, we present heterogeneous datapaths (HDPs), a novel system knob. Section 5.3 shows how multiple knobs can collaborate to deal with the variability of different timescales. Thereafter, the complete scheduling algorithm based on dynamic scenarios is elaborated in Sect. 5.4. Then, we further provide the proof that the algorithm can ensure timing guarantees in Sect. 5.5. Finally, the applicability of this approach to other architectures and technologies is discussed in Sect. 5.6.