The Scaling Behaviors in Achieving High Reliability via Chance-Constrained Optimization
Anand Deo, Karthyek Murthy
TL;DR
The paper develops a comprehensive scaling framework for chance-constrained optimization as reliability targets become extreme ($1-oldsymbol{ amealpha} o 1$). It proves that under mild regularity, the CCP's optimal value and decisions scale as $v_oldsymbol{ amealpha}^* o v^* s_oldsymbol{ amealpha}^r$ and $oldsymbol{x}_oldsymbol{ amealpha}^* o oldsymbol{x}^* s_oldsymbol{ amealpha}^r$, with the tail behavior of the marginals driving the rate via $s_oldsymbol{ amealpha}=ar{F}^{-1}(oldsymbol{ amealpha})$. The work then analyzes how distributional-robustifications (DRO with f-divergences, Wasserstein, and marginal-based sets) alter this scaling, showing that KL-divergence DRO can be exponentially conservative while marginal-based and certain f-divergences preserve scaling up to constants. It also develops methods to quantify and reduce conservativeness of CVaR and Bonferroni inner approximations and introduces a simple line-search to obtain asymptotically near-optimal CCP solutions. Finally, the authors propose EVT-inspired extrapolation to estimate Pareto-efficient decisions from limited data, providing a principled approach to handle $p(oldsymbol{x}) o 0$ regimes and to solve the P-Model with data-driven guarantees. The framework unifies risk-averse design across energy, cloud, and emergency-resource contexts, offering both theoretical scaling insights and practical algorithms for high-reliability decision-making under uncertainty.
Abstract
We study the problem of resource provisioning under stringent reliability or service level requirements, which arise in applications such as power distribution, emergency response, cloud server provisioning, and regulatory risk management. With chance-constrained optimization serving as a natural starting point for modeling this class of problems, our primary contribution is to characterize how the optimal costs and decisions scale for a generic joint chance-constrained model as the target probability of satisfying the service/reliability constraints approaches its maximal level. Beyond providing insights into the behavior of optimal solutions, our scaling framework has three key algorithmic implications. First, in distributionally robust optimization (DRO) modeling of chance constraints, we show that widely used approaches based on KL-divergences, Wasserstein distances, and moments heavily distort the scaling properties of optimal decisions, leading to exponentially higher costs. In contrast, incorporating marginal distributions or using appropriately chosen f-divergence balls preserves the correct scaling, ensuring decisions remain conservative by at most a constant or logarithmic factor. Second, we leverage the scaling framework to quantify the conservativeness of common inner approximations and propose a simple line search to refine their solutions, yielding near-optimal decisions. Finally, given N data samples, we demonstrate how the scaling framework enables the estimation of approximately Pareto-optimal decisions with constraint violation probabilities significantly smaller than the Omega(1/N)-barrier that arises in the absence of parametric assumptions