Engineering diffusive load balancing algorithms using experiments

Ralf Diekmann, Shanmugavelayutham Muthukrishnan, Madhu V. Nayakkankuppam

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    Abstract

    We study a distributed load balancing problem on arbitrary graphs. First Order (F0) and Second Order (SD) schemes are popular local diffusive schedules for this problem. To use them, several parameters have to be chosen carefully. Determining the "optimal" parameters analytically is difficult, and on a practical level, despite the widespread use of these schemes, little is known on how relevant parameters must be set. We employ systematic experiments to engineer the choice of relevant parameters in first and second order schemes. We present a centralized polynomial time algorithm for choosing the "optimal" F0 scheme based on sere[definite programming. Based on the empirical evidence from our implementation of this algorithm, we pose conjectures on the closed-form solution of optimal F0 schemes for various graphs. We also present a heuristic algorithm to locally estimate relevant parameters in the F0 and S0 schemes; our estimates are fairly accurate compared to those based on expensive global communication. Finally, we show that the F0 and S0 schemes that use approximate values rather than the optimal parameters, can be improved using a new iterative scheme that we introduce here; this scheme is of independent interest. The software we have developed for our implementations is available freely, and can serve as a platform for experimental research in this area. Our methods are being included in Paderborn, the Paderborn Finite Element Library [1].

    Original languageEnglish (US)
    Title of host publicationSolving Irregularly Structured Problems in Parallel - 4th International Symposium, IRREGULAR 1997, Proceedings
    EditorsReinhard Luling, Gianfranco Bilardi, Afonso Ferreira, Gianfranco Bilardi, Jose Rolim
    PublisherSpringer-Verlag
    Pages111-122
    Number of pages12
    ISBN (Print)3540631380, 9783540631385
    StatePublished - Jan 1 1997
    Event4th International Symposium on Solving Irregularly Structured Problems in Parallel, IRREGULAR 1997 - Paderborn, Germany
    Duration: Jun 12 1997Jun 13 1997

    Publication series

    NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
    Volume1253
    ISSN (Print)0302-9743
    ISSN (Electronic)1611-3349

    Conference

    Conference4th International Symposium on Solving Irregularly Structured Problems in Parallel, IRREGULAR 1997
    CountryGermany
    CityPaderborn
    Period6/12/976/13/97

    Fingerprint

    Load Balancing
    Resource allocation
    Engineering
    Heuristic algorithms
    Experiment
    Experiments
    Polynomials
    Engineers
    Communication
    Optimal Parameter
    First-order
    Graph in graph theory
    Iterative Scheme
    Closed-form Solution
    Heuristic algorithm
    Estimate
    Polynomial-time Algorithm
    Schedule
    Programming
    Finite Element

    ASJC Scopus subject areas

    • Theoretical Computer Science
    • Computer Science(all)

    Cite this

    Diekmann, R., Muthukrishnan, S., & Nayakkankuppam, M. V. (1997). Engineering diffusive load balancing algorithms using experiments. In R. Luling, G. Bilardi, A. Ferreira, G. Bilardi, & J. Rolim (Eds.), Solving Irregularly Structured Problems in Parallel - 4th International Symposium, IRREGULAR 1997, Proceedings (pp. 111-122). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1253). Springer-Verlag.

    Engineering diffusive load balancing algorithms using experiments. / Diekmann, Ralf; Muthukrishnan, Shanmugavelayutham; Nayakkankuppam, Madhu V.

    Solving Irregularly Structured Problems in Parallel - 4th International Symposium, IRREGULAR 1997, Proceedings. ed. / Reinhard Luling; Gianfranco Bilardi; Afonso Ferreira; Gianfranco Bilardi; Jose Rolim. Springer-Verlag, 1997. p. 111-122 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1253).

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    Diekmann, R, Muthukrishnan, S & Nayakkankuppam, MV 1997, Engineering diffusive load balancing algorithms using experiments. in R Luling, G Bilardi, A Ferreira, G Bilardi & J Rolim (eds), Solving Irregularly Structured Problems in Parallel - 4th International Symposium, IRREGULAR 1997, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 1253, Springer-Verlag, pp. 111-122, 4th International Symposium on Solving Irregularly Structured Problems in Parallel, IRREGULAR 1997, Paderborn, Germany, 6/12/97.
    Diekmann R, Muthukrishnan S, Nayakkankuppam MV. Engineering diffusive load balancing algorithms using experiments. In Luling R, Bilardi G, Ferreira A, Bilardi G, Rolim J, editors, Solving Irregularly Structured Problems in Parallel - 4th International Symposium, IRREGULAR 1997, Proceedings. Springer-Verlag. 1997. p. 111-122. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
    Diekmann, Ralf ; Muthukrishnan, Shanmugavelayutham ; Nayakkankuppam, Madhu V. / Engineering diffusive load balancing algorithms using experiments. Solving Irregularly Structured Problems in Parallel - 4th International Symposium, IRREGULAR 1997, Proceedings. editor / Reinhard Luling ; Gianfranco Bilardi ; Afonso Ferreira ; Gianfranco Bilardi ; Jose Rolim. Springer-Verlag, 1997. pp. 111-122 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
    @inproceedings{50f96e10e78f4c72b019e06d32f1a5d0,
    title = "Engineering diffusive load balancing algorithms using experiments",
    abstract = "We study a distributed load balancing problem on arbitrary graphs. First Order (F0) and Second Order (SD) schemes are popular local diffusive schedules for this problem. To use them, several parameters have to be chosen carefully. Determining the {"}optimal{"} parameters analytically is difficult, and on a practical level, despite the widespread use of these schemes, little is known on how relevant parameters must be set. We employ systematic experiments to engineer the choice of relevant parameters in first and second order schemes. We present a centralized polynomial time algorithm for choosing the {"}optimal{"} F0 scheme based on sere[definite programming. Based on the empirical evidence from our implementation of this algorithm, we pose conjectures on the closed-form solution of optimal F0 schemes for various graphs. We also present a heuristic algorithm to locally estimate relevant parameters in the F0 and S0 schemes; our estimates are fairly accurate compared to those based on expensive global communication. Finally, we show that the F0 and S0 schemes that use approximate values rather than the optimal parameters, can be improved using a new iterative scheme that we introduce here; this scheme is of independent interest. The software we have developed for our implementations is available freely, and can serve as a platform for experimental research in this area. Our methods are being included in Paderborn, the Paderborn Finite Element Library [1].",
    author = "Ralf Diekmann and Shanmugavelayutham Muthukrishnan and Nayakkankuppam, {Madhu V.}",
    year = "1997",
    month = "1",
    day = "1",
    language = "English (US)",
    isbn = "3540631380",
    series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
    publisher = "Springer-Verlag",
    pages = "111--122",
    editor = "Reinhard Luling and Gianfranco Bilardi and Afonso Ferreira and Gianfranco Bilardi and Jose Rolim",
    booktitle = "Solving Irregularly Structured Problems in Parallel - 4th International Symposium, IRREGULAR 1997, Proceedings",

    }

    TY - GEN

    T1 - Engineering diffusive load balancing algorithms using experiments

    AU - Diekmann, Ralf

    AU - Muthukrishnan, Shanmugavelayutham

    AU - Nayakkankuppam, Madhu V.

    PY - 1997/1/1

    Y1 - 1997/1/1

    N2 - We study a distributed load balancing problem on arbitrary graphs. First Order (F0) and Second Order (SD) schemes are popular local diffusive schedules for this problem. To use them, several parameters have to be chosen carefully. Determining the "optimal" parameters analytically is difficult, and on a practical level, despite the widespread use of these schemes, little is known on how relevant parameters must be set. We employ systematic experiments to engineer the choice of relevant parameters in first and second order schemes. We present a centralized polynomial time algorithm for choosing the "optimal" F0 scheme based on sere[definite programming. Based on the empirical evidence from our implementation of this algorithm, we pose conjectures on the closed-form solution of optimal F0 schemes for various graphs. We also present a heuristic algorithm to locally estimate relevant parameters in the F0 and S0 schemes; our estimates are fairly accurate compared to those based on expensive global communication. Finally, we show that the F0 and S0 schemes that use approximate values rather than the optimal parameters, can be improved using a new iterative scheme that we introduce here; this scheme is of independent interest. The software we have developed for our implementations is available freely, and can serve as a platform for experimental research in this area. Our methods are being included in Paderborn, the Paderborn Finite Element Library [1].

    AB - We study a distributed load balancing problem on arbitrary graphs. First Order (F0) and Second Order (SD) schemes are popular local diffusive schedules for this problem. To use them, several parameters have to be chosen carefully. Determining the "optimal" parameters analytically is difficult, and on a practical level, despite the widespread use of these schemes, little is known on how relevant parameters must be set. We employ systematic experiments to engineer the choice of relevant parameters in first and second order schemes. We present a centralized polynomial time algorithm for choosing the "optimal" F0 scheme based on sere[definite programming. Based on the empirical evidence from our implementation of this algorithm, we pose conjectures on the closed-form solution of optimal F0 schemes for various graphs. We also present a heuristic algorithm to locally estimate relevant parameters in the F0 and S0 schemes; our estimates are fairly accurate compared to those based on expensive global communication. Finally, we show that the F0 and S0 schemes that use approximate values rather than the optimal parameters, can be improved using a new iterative scheme that we introduce here; this scheme is of independent interest. The software we have developed for our implementations is available freely, and can serve as a platform for experimental research in this area. Our methods are being included in Paderborn, the Paderborn Finite Element Library [1].

    UR - http://www.scopus.com/inward/record.url?scp=84947796534&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=84947796534&partnerID=8YFLogxK

    M3 - Conference contribution

    AN - SCOPUS:84947796534

    SN - 3540631380

    SN - 9783540631385

    T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

    SP - 111

    EP - 122

    BT - Solving Irregularly Structured Problems in Parallel - 4th International Symposium, IRREGULAR 1997, Proceedings

    A2 - Luling, Reinhard

    A2 - Bilardi, Gianfranco

    A2 - Ferreira, Afonso

    A2 - Bilardi, Gianfranco

    A2 - Rolim, Jose

    PB - Springer-Verlag

    ER -