Hyper-scalable JSQ with sparse feedback

Onderzoeksoutput: Hoofdstuk in Boek/Rapport/CongresprocedureConferentiebijdrageAcademicpeer review

Uittreksel

Load balancing algorithms play a vital role in enhancing performance in data centers and cloud networks. Due to the massive size of these systems, scalability challenges, and especially the communication overhead associated with load balancing mechanisms, have emerged as major concerns. Motivated by these issues, we introduce and analyze a novel class of load balancing schemes where the various servers provide occasional queue updates to guide the load assignment. We show that the proposed schemes strongly outperform JSQ(d) strategies with comparable communication overhead per job, and can achieve a vanishing waiting time in the many-server limit with just one message per job, just like the popular JIQ scheme. The proposed schemes are particularly geared however towards the sparse feedback regime with less than one message per job, where they outperform corresponding sparsified JIQ versions. We investigate fluid limits for synchronous updates as well as asynchronous exponential update intervals. The fixed point of the fluid limit is identified in the latter case, and used to derive the queue length distribution. We also demonstrate that in the ultralow feedback regime the mean stationary waiting time tends to a constant in the synchronous case, but grows without bound in the asynchronous case.

TaalEngels
TitelSIGMETRICS Performance 2019 - Abstracts of the 2019 SIGMETRICS/Performance Joint International Conference on Measurement and Modeling of Computer Systems
Plaats van productieNew York
UitgeverijAssociation for Computing Machinery, Inc
Pagina's61-62
Aantal pagina's2
ISBN van elektronische versie978-1-4503-6678-6
DOI's
StatusGepubliceerd - 20 jun 2019
Evenement14th Joint Conference of International Conference on Measurement and Modeling of Computer Systems, SIGMETRICS 2019 and IFIP Performance Conference 2019, SIGMETRICS/Performance 2019 - Phoenix, Verenigde Staten van Amerika
Duur: 24 jun 201928 jun 2019

Congres

Congres14th Joint Conference of International Conference on Measurement and Modeling of Computer Systems, SIGMETRICS 2019 and IFIP Performance Conference 2019, SIGMETRICS/Performance 2019
LandVerenigde Staten van Amerika
StadPhoenix
Periode24/06/1928/06/19

Vingerafdruk

Resource allocation
Feedback
Servers
Fluids
Communication
Scalability

Trefwoorden

    Citeer dit

    van der Boor, M., Borst, S., & van Leeuwaarden, J. (2019). Hyper-scalable JSQ with sparse feedback. In SIGMETRICS Performance 2019 - Abstracts of the 2019 SIGMETRICS/Performance Joint International Conference on Measurement and Modeling of Computer Systems (blz. 61-62). New York: Association for Computing Machinery, Inc. DOI: 10.1145/3309697.3331477
    van der Boor, Mark ; Borst, Sem ; van Leeuwaarden, Johan. / Hyper-scalable JSQ with sparse feedback. SIGMETRICS Performance 2019 - Abstracts of the 2019 SIGMETRICS/Performance Joint International Conference on Measurement and Modeling of Computer Systems. New York : Association for Computing Machinery, Inc, 2019. blz. 61-62
    @inproceedings{9a4e9304450747a586dcaa1d6028e68d,
    title = "Hyper-scalable JSQ with sparse feedback",
    abstract = "Load balancing algorithms play a vital role in enhancing performance in data centers and cloud networks. Due to the massive size of these systems, scalability challenges, and especially the communication overhead associated with load balancing mechanisms, have emerged as major concerns. Motivated by these issues, we introduce and analyze a novel class of load balancing schemes where the various servers provide occasional queue updates to guide the load assignment. We show that the proposed schemes strongly outperform JSQ(d) strategies with comparable communication overhead per job, and can achieve a vanishing waiting time in the many-server limit with just one message per job, just like the popular JIQ scheme. The proposed schemes are particularly geared however towards the sparse feedback regime with less than one message per job, where they outperform corresponding sparsified JIQ versions. We investigate fluid limits for synchronous updates as well as asynchronous exponential update intervals. The fixed point of the fluid limit is identified in the latter case, and used to derive the queue length distribution. We also demonstrate that in the ultralow feedback regime the mean stationary waiting time tends to a constant in the synchronous case, but grows without bound in the asynchronous case.",
    keywords = "Cloud networks, Data centers, Delay performance, Join-the-shortest-queue, Load balancing, Parallelserver systems, Scaling limits",
    author = "{van der Boor}, Mark and Sem Borst and {van Leeuwaarden}, Johan",
    year = "2019",
    month = "6",
    day = "20",
    doi = "10.1145/3309697.3331477",
    language = "English",
    pages = "61--62",
    booktitle = "SIGMETRICS Performance 2019 - Abstracts of the 2019 SIGMETRICS/Performance Joint International Conference on Measurement and Modeling of Computer Systems",
    publisher = "Association for Computing Machinery, Inc",
    address = "United States",

    }

    van der Boor, M, Borst, S & van Leeuwaarden, J 2019, Hyper-scalable JSQ with sparse feedback. in SIGMETRICS Performance 2019 - Abstracts of the 2019 SIGMETRICS/Performance Joint International Conference on Measurement and Modeling of Computer Systems. Association for Computing Machinery, Inc, New York, blz. 61-62, Phoenix, Verenigde Staten van Amerika, 24/06/19. DOI: 10.1145/3309697.3331477

    Hyper-scalable JSQ with sparse feedback. / van der Boor, Mark; Borst, Sem; van Leeuwaarden, Johan.

    SIGMETRICS Performance 2019 - Abstracts of the 2019 SIGMETRICS/Performance Joint International Conference on Measurement and Modeling of Computer Systems. New York : Association for Computing Machinery, Inc, 2019. blz. 61-62.

    Onderzoeksoutput: Hoofdstuk in Boek/Rapport/CongresprocedureConferentiebijdrageAcademicpeer review

    TY - GEN

    T1 - Hyper-scalable JSQ with sparse feedback

    AU - van der Boor,Mark

    AU - Borst,Sem

    AU - van Leeuwaarden,Johan

    PY - 2019/6/20

    Y1 - 2019/6/20

    N2 - Load balancing algorithms play a vital role in enhancing performance in data centers and cloud networks. Due to the massive size of these systems, scalability challenges, and especially the communication overhead associated with load balancing mechanisms, have emerged as major concerns. Motivated by these issues, we introduce and analyze a novel class of load balancing schemes where the various servers provide occasional queue updates to guide the load assignment. We show that the proposed schemes strongly outperform JSQ(d) strategies with comparable communication overhead per job, and can achieve a vanishing waiting time in the many-server limit with just one message per job, just like the popular JIQ scheme. The proposed schemes are particularly geared however towards the sparse feedback regime with less than one message per job, where they outperform corresponding sparsified JIQ versions. We investigate fluid limits for synchronous updates as well as asynchronous exponential update intervals. The fixed point of the fluid limit is identified in the latter case, and used to derive the queue length distribution. We also demonstrate that in the ultralow feedback regime the mean stationary waiting time tends to a constant in the synchronous case, but grows without bound in the asynchronous case.

    AB - Load balancing algorithms play a vital role in enhancing performance in data centers and cloud networks. Due to the massive size of these systems, scalability challenges, and especially the communication overhead associated with load balancing mechanisms, have emerged as major concerns. Motivated by these issues, we introduce and analyze a novel class of load balancing schemes where the various servers provide occasional queue updates to guide the load assignment. We show that the proposed schemes strongly outperform JSQ(d) strategies with comparable communication overhead per job, and can achieve a vanishing waiting time in the many-server limit with just one message per job, just like the popular JIQ scheme. The proposed schemes are particularly geared however towards the sparse feedback regime with less than one message per job, where they outperform corresponding sparsified JIQ versions. We investigate fluid limits for synchronous updates as well as asynchronous exponential update intervals. The fixed point of the fluid limit is identified in the latter case, and used to derive the queue length distribution. We also demonstrate that in the ultralow feedback regime the mean stationary waiting time tends to a constant in the synchronous case, but grows without bound in the asynchronous case.

    KW - Cloud networks

    KW - Data centers

    KW - Delay performance

    KW - Join-the-shortest-queue

    KW - Load balancing

    KW - Parallelserver systems

    KW - Scaling limits

    UR - http://www.scopus.com/inward/record.url?scp=85069184635&partnerID=8YFLogxK

    U2 - 10.1145/3309697.3331477

    DO - 10.1145/3309697.3331477

    M3 - Conference contribution

    SP - 61

    EP - 62

    BT - SIGMETRICS Performance 2019 - Abstracts of the 2019 SIGMETRICS/Performance Joint International Conference on Measurement and Modeling of Computer Systems

    PB - Association for Computing Machinery, Inc

    CY - New York

    ER -

    van der Boor M, Borst S, van Leeuwaarden J. Hyper-scalable JSQ with sparse feedback. In SIGMETRICS Performance 2019 - Abstracts of the 2019 SIGMETRICS/Performance Joint International Conference on Measurement and Modeling of Computer Systems. New York: Association for Computing Machinery, Inc. 2019. blz. 61-62. Beschikbaar vanaf, DOI: 10.1145/3309697.3331477