Multi-Objectivizing Software Configuration Tuning (for a single performance concern)
This addresses a key bottleneck in automated software optimization for developers and engineers, offering a novel approach to improve efficiency in expensive measurement settings.
The paper tackles the problem of software configuration tuning being trapped in local optima by proposing a meta multi-objectivization model that uses an auxiliary performance objective to make similar configurations less comparable, preventing search stagnation. Experiments on eight real-world systems show it is statistically more effective than state-of-the-art single-objective methods, with up to 42% gain and using as low as 24% of measurements to achieve the same or better performance.
Automatically tuning software configuration for optimizing a single performance attribute (e.g., minimizing latency) is not trivial, due to the nature of the configuration systems (e.g., complex landscape and expensive measurement). To deal with the problem, existing work has been focusing on developing various effective optimizers. However, a prominent issue that all these optimizers need to take care of is how to avoid the search being trapped in local optima -- a hard nut to crack for software configuration tuning due to its rugged and sparse landscape, and neighboring configurations tending to behave very differently. Overcoming such in an expensive measurement setting is even more challenging. In this paper, we take a different perspective to tackle this issue. Instead of focusing on improving the optimizer, we work on the level of optimization model. We do this by proposing a meta multi-objectivization model (MMO) that considers an auxiliary performance objective (e.g., throughput in addition to latency). What makes this model unique is that we do not optimize the auxiliary performance objective, but rather use it to make similarly-performing while different configurations less comparable (i.e. Pareto nondominated to each other), thus preventing the search from being trapped in local optima. Experiments on eight real-world software systems/environments with diverse performance attributes reveal that our MMO model is statistically more effective than state-of-the-art single-objective counterparts in overcoming local optima (up to 42% gain), while using as low as 24% of their measurements to achieve the same (or better) performance result.