🤖 AI Summary
This paper studies the bilateral online resource allocation problem with cancellation costs, focusing on edge-weighted online bipartite matching and its extensions—including deterministic integral allocations—where allocated resources may be bought back at a cost comprising the original edge weight plus an additional penalty proportional to the weight with factor $f geq 0$. We provide the first complete characterization of the optimal competitive ratio for arbitrary $f$, revealing a phase transition between small and large $f$, and establish a unified analytical framework. Our algorithm is designed via a parameterized primal-dual approach, integrating closed-form analysis using the Lambert $W$ function and piecewise lower-bound constructions. The resulting tight competitive ratios are: $max{e/(e-1-f),, 2/(1-f)}$ for small $f$, and either $-W_{-1}(-1/(e(1+f)))$ or $1+2f+2sqrt{f(1+f)}$ for large $f$, with matching upper and lower bounds.
📝 Abstract
Motivated by applications in cloud computing spot markets and selling banner ads on popular websites, we study the online resource allocation problem with overbooking and cancellation costs, also known as the emph{buyback} setting. To model this problem, we consider a variation of the classic edge-weighted online matching problem in which the decision maker can reclaim any fraction of an offline resource that is pre-allocated to an earlier online vertex; however, by doing so not only the decision maker loses the previously allocated edge-weight, it also has to pay a non-negative constant factor $f$ of this edge-weight as an extra penalty. Parameterizing the problem by the buyback factor $f$, our main result is obtaining optimal competitive algorithms for emph{all possible values} of $f$ through a novel primal-dual family of algorithms. We establish the optimality of our results by obtaining separate lower bounds for each of small and large buyback factor regimes and showing how our primal-dual algorithm exactly matches this lower bound by appropriately tuning a parameter as a function of $f$. Interestingly, our result shows a phase transition: for small buyback regime ($f<frac{e-2}{2}$), the optimal competitive ratio is $frac{e}{e-(1+f)}$, and for the large buyback regime ($fgeq frac{e-2}{2}$), the competitive ratio is $-W_{-1}left(frac{-1}{e(1+f)}
ight)$, where $W_{-1}$ is the non-principal branch of the Lambert W function. We further study the optimal competitive ratio in variants of this model using our unifying framework, such as matching with deterministic integral allocations or single-resource with different demands. For deterministic integral matching, our results again show a phase transition: for small buyback regime ($f<frac{1}{3}$), the optimal competitive ratio is $frac{2}{1-f}$, and for the large buyback regime ($fgeq frac{1}{3}$), it is $1+2f+2sqrt{f(1+f)}$.
