VEER: enhancing the interpretability of model-based optimizations

Many software systems can be tuned for multiple objectives (e.g., faster runtime, less required memory, network traffic or energy consumption, etc.). Such suffer from “disagreement” where different models have (or even opposite) insights and tactics on how to optimize a system. For configuration problems, we show that (a) model disagreement is rampant; yet (b) prior this paper, it has barely been explored. We aim at helping practitioners researchers better solve multi-objective optimization by resolving disagreement. propose dimension reduction method called VEER builds useful one-dimensional approximation the original N-objective space. Traditional model-based optimizers use Pareto search locate Pareto-optimal solutions problem, which computationally heavy large-scale systems. surrogate replace sorting step after deployment. Compared state-of-the-art, 11 configurable systems, significantly reduces execution time, without compromising performance in most cases. our largest problem (with tens of thousands possible configurations), optimizing with finds as good optimizations zero disagreements, three orders magnitude faster. When employing optimization, recommend apply VEER, not only improves but also resolves potential problem.