PulseTrendingJournal ClubResearchersJournalsExplore
Instagram
HomeTrendingJournal ClubExplore
Synapse
⌘+K
Synapse
August 28, 2026Manufacturing & Service Operations ManagementOpen Access

Service Systems with On-Demand and Reserved Servers

View Full Paper
Ask AI
Bookmark
Share

Authors

ZFZhichao FengMDMilind DawandeGJGanesh Janakiraman

Discussion

Loading...

Member takes

Overview

Queueing analysis demonstrates that a threshold-based bang-bang policy minimizes server and waiting costs in hybrid service systems, indicating optimal strategies for dynamic cloud provisioning.

Key Points

  • To determine the optimal dynamic allocation and capacity sizing that minimizes server hiring and customer waiting costs in queueing systems using both reserved and on-demand servers.
  • Formulated a dynamic optimization queueing model combining upfront reserved server capacity with real-time on-demand server hiring under waiting and hiring costs.
  • Developed algorithmic frameworks targeting discounted-cost criteria and average-cost objectives.
  • Extended the formulation to account for operational frictions, specifically fixed switching costs and provisioning latency, utilizing a periodic tracking policy.
  • Proved that optimal on-demand hiring follows a threshold-based bang-bang policy: zero on-demand servers operate below the job threshold, whereas above it, sufficient servers deploy to clear all waiting jobs.
  • Demonstrated the asymptotic optimality of the periodic tracking policy under operational frictions including switching costs and provisioning delays.
  • Provided prescriptive guidelines through numerical experiments showing how reserved baseline capacity and on-demand capacity scale with job urgency and arrival volume.

Cite This Study

Feng et al. (2026) studied this question.

synapsesocial.com/papers/6a914599d15324a1df3a8dbbhttps://doi.org/10.1287/msom.2024.1504
View Full Paper
Ask AI
Bookmark
Share