The "w" network and the dynamic control of unreliable flexible servers

Soroush Saghafian, Mark P. Van Oyen, Bora Kolfal

Research output: Contribution to journalArticle

12 Citations (Scopus)

Abstract

This article addresses the problem of effectively assigning partially flexible resources to various jobs in Markovian parallel queueing systems with heterogeneous and unreliable servers. Attention is focused on a structure forming a W and it is found that this design is highly efficient; it requires only a small amount of cross-training but often performs almost as well as a fully cross-trained system. It is shown that (even allowing disruptions) a version of the c rule, which prioritizes serving the fixed task before the shared, is optimal under some conditions. Since the optimal policy is complex in general, a powerful and yet simple control policy is developed. This policy (which is implementable in any parallel queueing system) defines a simple measure of workload costs and assigns each server to the queue with the Largest Expected Workload Cost (LEWC). Thus, it effectively combines the intuition underlying two widely used policies: (i) the load-balancing objective in serving the Longest Queue (LQ); and (ii) the greedy cost minimization emphasis of the c rule. Extensive numerical tests show that LEWC performs well in comparison with four key policies: optimal, LQ, c, and generalized c (Gc). The stability of the LEWC, LQ, and Gc policies is proved.

Original languageEnglish (US)
Pages (from-to)893-907
Number of pages15
JournalIIE Transactions (Institute of Industrial Engineers)
Volume43
Issue number12
DOIs
StatePublished - Dec 2011
Externally publishedYes

Fingerprint

Servers
Costs
Resource allocation

Keywords

  • control of queues
  • Flexible servers
  • Markov decision process
  • stochastic resource allocation
  • unreliable servers

ASJC Scopus subject areas

  • Industrial and Manufacturing Engineering

Cite this

The "w" network and the dynamic control of unreliable flexible servers. / Saghafian, Soroush; Van Oyen, Mark P.; Kolfal, Bora.

In: IIE Transactions (Institute of Industrial Engineers), Vol. 43, No. 12, 12.2011, p. 893-907.

Research output: Contribution to journalArticle

Saghafian, Soroush ; Van Oyen, Mark P. ; Kolfal, Bora. / The "w" network and the dynamic control of unreliable flexible servers. In: IIE Transactions (Institute of Industrial Engineers). 2011 ; Vol. 43, No. 12. pp. 893-907.
@article{910e0ac8eed6403494fc08999f7f3561,
title = "The {"}w{"} network and the dynamic control of unreliable flexible servers",
abstract = "This article addresses the problem of effectively assigning partially flexible resources to various jobs in Markovian parallel queueing systems with heterogeneous and unreliable servers. Attention is focused on a structure forming a W and it is found that this design is highly efficient; it requires only a small amount of cross-training but often performs almost as well as a fully cross-trained system. It is shown that (even allowing disruptions) a version of the c rule, which prioritizes serving the fixed task before the shared, is optimal under some conditions. Since the optimal policy is complex in general, a powerful and yet simple control policy is developed. This policy (which is implementable in any parallel queueing system) defines a simple measure of workload costs and assigns each server to the queue with the Largest Expected Workload Cost (LEWC). Thus, it effectively combines the intuition underlying two widely used policies: (i) the load-balancing objective in serving the Longest Queue (LQ); and (ii) the greedy cost minimization emphasis of the c rule. Extensive numerical tests show that LEWC performs well in comparison with four key policies: optimal, LQ, c, and generalized c (Gc). The stability of the LEWC, LQ, and Gc policies is proved.",
keywords = "control of queues, Flexible servers, Markov decision process, stochastic resource allocation, unreliable servers",
author = "Soroush Saghafian and {Van Oyen}, {Mark P.} and Bora Kolfal",
year = "2011",
month = "12",
doi = "10.1080/0740817X.2011.575678",
language = "English (US)",
volume = "43",
pages = "893--907",
journal = "IISE Transactions",
issn = "2472-5854",
publisher = "Taylor and Francis Ltd.",
number = "12",

}

TY - JOUR

T1 - The "w" network and the dynamic control of unreliable flexible servers

AU - Saghafian, Soroush

AU - Van Oyen, Mark P.

AU - Kolfal, Bora

PY - 2011/12

Y1 - 2011/12

N2 - This article addresses the problem of effectively assigning partially flexible resources to various jobs in Markovian parallel queueing systems with heterogeneous and unreliable servers. Attention is focused on a structure forming a W and it is found that this design is highly efficient; it requires only a small amount of cross-training but often performs almost as well as a fully cross-trained system. It is shown that (even allowing disruptions) a version of the c rule, which prioritizes serving the fixed task before the shared, is optimal under some conditions. Since the optimal policy is complex in general, a powerful and yet simple control policy is developed. This policy (which is implementable in any parallel queueing system) defines a simple measure of workload costs and assigns each server to the queue with the Largest Expected Workload Cost (LEWC). Thus, it effectively combines the intuition underlying two widely used policies: (i) the load-balancing objective in serving the Longest Queue (LQ); and (ii) the greedy cost minimization emphasis of the c rule. Extensive numerical tests show that LEWC performs well in comparison with four key policies: optimal, LQ, c, and generalized c (Gc). The stability of the LEWC, LQ, and Gc policies is proved.

AB - This article addresses the problem of effectively assigning partially flexible resources to various jobs in Markovian parallel queueing systems with heterogeneous and unreliable servers. Attention is focused on a structure forming a W and it is found that this design is highly efficient; it requires only a small amount of cross-training but often performs almost as well as a fully cross-trained system. It is shown that (even allowing disruptions) a version of the c rule, which prioritizes serving the fixed task before the shared, is optimal under some conditions. Since the optimal policy is complex in general, a powerful and yet simple control policy is developed. This policy (which is implementable in any parallel queueing system) defines a simple measure of workload costs and assigns each server to the queue with the Largest Expected Workload Cost (LEWC). Thus, it effectively combines the intuition underlying two widely used policies: (i) the load-balancing objective in serving the Longest Queue (LQ); and (ii) the greedy cost minimization emphasis of the c rule. Extensive numerical tests show that LEWC performs well in comparison with four key policies: optimal, LQ, c, and generalized c (Gc). The stability of the LEWC, LQ, and Gc policies is proved.

KW - control of queues

KW - Flexible servers

KW - Markov decision process

KW - stochastic resource allocation

KW - unreliable servers

UR - http://www.scopus.com/inward/record.url?scp=80053555732&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=80053555732&partnerID=8YFLogxK

U2 - 10.1080/0740817X.2011.575678

DO - 10.1080/0740817X.2011.575678

M3 - Article

VL - 43

SP - 893

EP - 907

JO - IISE Transactions

JF - IISE Transactions

SN - 2472-5854

IS - 12

ER -