QKnober: A Knob-Based Fairness-Efficiency Scheduler for Cloud Computing with QoS Guarantees

Abstract

Fairness and efficiency are generally two important metrics for users in modern cloud computing. Due to the heterogeneous resource demands of CPU and memory for users’ tasks, it cannot achieve the strict 100% fairness and the maximum efficiency at the same time. Quantitatively showing the fairness degradation/loss becomes essentially important in the design of any fairness-efficiency tradeoff scheduler. Existing fairness-efficiency schedulers (e.g., Tetris) can balance such a tradeoff elastically by relaxing fairness constraint for improved efficiency using the knob. However, their approaches are insensitive to the fairness degradation under different knobs, which makes several drawbacks. First, it cannot quantitatively tell how much relaxed fairness can be guaranteed (i.e., QoS of fairness guarantee) given a knob value. Second, it fails to meet several essential properties such as sharing incentive. To address these issues, we propose a new fairness-efficiency scheduler, QKnober, to balance the fairness and efficiency elastically and flexibly using a tunable fairness knob. QKnober is a fairness-sensitive scheduler that can maximize the system efficiency while guaranteeing the \(\theta \)-soft fairness by modeling the whole allocation as a combination of fairness-purpose allocation and efficiency-purpose allocation. Moreover, QKnober satisfies fairness properties of sharing incentive, envy-freeness and pareto efficiency given a proper knob. We have implemented QKnober in YARN and evaluated it using real experiments. The results show that QKnober can achieve good performance and fairness.