Social Cost Analysis of Shared/Buy-in Computing Systems

Zhenpeng Shi, David Starobinski, Ariel Orda

Research output: Contribution to journalArticlepeer-review

Abstract

Shared/buy-in computing systems offer users the option to select between buy-in and shared services. In such systems, idle buy-in resources are made available to other users for sharing. With strategic users, resource purchase and allocation in such systems can be cast as a non-cooperative game, whose corresponding Nash equilibrium does not necessarily result in the optimal social cost. In this study, we first derive the optimal social cost of the game in closed form, by casting it as a convex optimization problem and establishing related properties. Next, we derive a closed-form expression for the social cost at the Nash equilibrium, and show that it can be computed in linear time. We further show that the strategy profiles of users at the optimum and the Nash equilibrium are directly proportional. We measure the inefficiency of the Nash equilibrium through the price of anarchy, and show that it can be quite large in certain cases, e.g., when the operating expense ratio is low or when the distribution of user workloads is relatively homogeneous. To improve the efficiency of the system, we propose and analyze two subsidy policies, which are shown to converge using best-response dynamics.

Original languageEnglish
Article number9
JournalACM Transactions on Economics and Computation
Volume11
Issue number3-4
DOIs
StatePublished - 19 Dec 2023

Keywords

  • Computing clusters
  • aggregative games
  • efficiency
  • price of anarchy
  • pricing
  • social welfare

All Science Journal Classification (ASJC) codes

  • Computer Science (miscellaneous)
  • Computational Mathematics
  • Economics and Econometrics
  • Marketing
  • Statistics and Probability

Fingerprint

Dive into the research topics of 'Social Cost Analysis of Shared/Buy-in Computing Systems'. Together they form a unique fingerprint.

Cite this