Distributed allocation of computing tasks over network resources is meant to decrease the cost of centralized allocation. However, existing analytical models consider practically indistinguishable resources, e.g., located in the data center. With the rise of edge computing, it becomes important to account for the impact of diverse latency values imposed by edge/cloud data center locations. In this paper, we study the optimization of computing task allocation considering both the delays to reach edge/cloud data centers and the response times of servers. We explicitly evaluate the resulting performance under different scenarios. We show, through numerical analysis and real experiments, that differences in delays to reach data center locations cannot be neglected. We also study the price of anarchy of a distributed implementation of the computing task allocation and unveil important properties such as the price of anarchy being generally small, except when the system is overloaded, and its maximum can be computed with low complexity.

Optimal Allocation of Tasks to Networked Computing Facilities

Castagno P.;Sereno M.;
2025-01-01

Abstract

Distributed allocation of computing tasks over network resources is meant to decrease the cost of centralized allocation. However, existing analytical models consider practically indistinguishable resources, e.g., located in the data center. With the rise of edge computing, it becomes important to account for the impact of diverse latency values imposed by edge/cloud data center locations. In this paper, we study the optimization of computing task allocation considering both the delays to reach edge/cloud data centers and the response times of servers. We explicitly evaluate the resulting performance under different scenarios. We show, through numerical analysis and real experiments, that differences in delays to reach data center locations cannot be neglected. We also study the price of anarchy of a distributed implementation of the computing task allocation and unveil important properties such as the price of anarchy being generally small, except when the system is overloaded, and its maximum can be computed with low complexity.
2025
28th International Conference on Analytical and Stochastic Modeling Techniques and Applications, ASMTA 2024
ita
2024
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Springer Science and Business Media Deutschland GmbH
14826
33
50
9783031707520
9783031707537
Game Theory; Network servers; Next generation networking; Optimization with network latency constraints; Price of Anarchy
Mancuso V.; Castagno P.; Badia L.; Sereno M.; Ajmone Marsan M.
File in questo prodotto:
File Dimensione Formato  
McCloud_meets_Pigou.pdf

Accesso aperto con embargo fino al 01/01/2027

Tipo di file: PREPRINT (PRIMA BOZZA)
Dimensione 589.06 kB
Formato Adobe PDF
589.06 kB Adobe PDF   Visualizza/Apri   Richiedi una copia

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/2318/2080330
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? 0
social impact