[Paper Review] Stability of Service under Time-of-Use Pricing

Shuchi Chawla, Nikhil R. Devanur+4 more|arXiv (Cornell University)|Apr 7, 2017

Optimization and Search Problems16 references4 citations

TL;DR

This paper proposes time-of-use pricing as a mechanism to stabilize resource allocation in stochastic temporal markets, such as energy or cloud computing, where demand fluctuates. By setting prices based on expected demand, the system achieves near-optimal social welfare with high probability, even under adversarial job arrival orders and stochastic overload risks, due to a carefully constructed forwarding graph and capacity reserves that ensure robustness to demand variability.

ABSTRACT

We consider "time-of-use" pricing as a technique for matching supply and demand of temporal resources with the goal of maximizing social welfare. Relevant examples include energy, computing resources on a cloud computing platform, and charging stations for electric vehicles, among many others. A client/job in this setting has a window of time during which he needs service, and a particular value for obtaining it. We assume a stochastic model for demand, where each job materializes with some probability via an independent Bernoulli trial. Given a per-time-unit pricing of resources, any realized job will first try to get served by the cheapest available resource in its window and, failing that, will try to find service at the next cheapest available resource, and so on. Thus, the natural stochastic fluctuations in demand have the potential to lead to cascading overload events. Our main result shows that setting prices so as to optimally handle the {\em expected} demand works well: with high probability, when the actual demand is instantiated, the system is stable and the expected value of the jobs served is very close to that of the optimal offline algorithm.

Motivation & Objective

• To design a pricing mechanism that stabilizes supply-demand matching in temporal resource systems with stochastic demand fluctuations.
• To ensure high-probability stability and near-optimal social welfare despite adversarial job arrival orders and random demand realizations.
• To analyze the impact of demand variability on cascading overload events and show that optimal expected pricing mitigates these risks.
• To prove that setting prices based on expected demand, with modest overprovisioning, yields near-optimal performance in high-probability scenarios.
• To establish truthfulness and efficiency of the mechanism under realistic assumptions of job behavior and resource constraints.

Proposed method

• The authors model job arrivals as independent Bernoulli trials with given probabilities, and define a forwarding graph where edges represent potential demand forwarding from one time slot to a more expensive one.
• They introduce a layered network structure over time blocks, partitioning the timeline into blocks of varying lengths to model job durations and price dependencies.
• A network ${\mathcal{D}}$ is constructed with intra-layer edges (forwarding and backward) and inter-layer edges connecting blocks of the same time slot across lengths, forming a graph with maximum in-degree 4.
• The mechanism uses a price vector $p_t(l)$ per time slot and block length $l$, ensuring that jobs prefer the cheapest available slot within their window and duration.
• The analysis leverages moment generating functions and concentration inequalities to bound the probability of overload, showing that the expected number of jobs served is close to optimal.
• A short-cutting technique is applied to realized job paths, transforming them into paths in the layered graph ${\mathcal{D}}$, enabling application of a min-work condition and probabilistic analysis.

Experimental results

Research questions

• RQ1 Can time-of-use pricing based on expected demand achieve high-probability stability and near-optimal social welfare in stochastic temporal resource allocation?
• RQ2 How do stochastic demand fluctuations and adversarial job arrival orders affect system stability and performance?
• RQ3 What structural properties of the forwarding graph determine the likelihood of cascading overload events?
• RQ4 Can a mechanism that sets prices based on expected demand still guarantee near-optimal performance without increasing capacity beyond single-period requirements?
• RQ5 Is the mechanism truthful with respect to job values, windows, and lengths under the proposed pricing and allocation rules?

Key findings

• Setting prices based on expected demand ensures that the expected value of jobs served is at least $1 - \epsilon$ times the optimal offline solution, with high probability.
• The system remains stable under adversarial job arrival order and stochastic demand, provided that capacity $B_t = \Omega(\epsilon^{-2} \ln(\epsilon^{-1}))$.
• The mechanism achieves a $1 - 2\epsilon$ approximation to social welfare, with high probability, due to bounded moment generating functions of excess demand.
• The forwarding graph’s maximum in-degree is 4, which enables the use of a min-work condition and probabilistic concentration bounds.
• The mechanism is truthful: jobs cannot benefit by misreporting their value, window, or length, as they are allocated the cheapest available slot at arrival.
• The analysis holds even when jobs are allowed to purchase longer-duration blocks at the price of their true length, with only minor modifications to the mechanism.