Tackling latency via replication in distributed systems

Título traducido de la contribución: Tratamiento de la latencia mediante replicación en sistemas distribuidos

Zhan Qiu, Juan F. Pérez, Peter G. Harrison

Resultado de la investigación: Contribución a libro /Tipo informe o reporteContribución en conferencia

4 Citas (Scopus)

Resumen

Consistently high reliability and low latency are twin requirements common to many forms of distributed processing; for example, server farms and mirrored storage access. To address them, we consider replication of requests with canceling – i.e. initiate multiple concurrent replicas of a request and use the first successful result returned, canceling all outstanding replicas. This scheme has been studied recently, but mostly for systems with a single central queue, while server farms exploit distributed resources for scalability and robustness. We develop an approximate stochastic model to determine the response time distribution in a system with distributed queues, and compare its performance against its centralized counterpart. Validation against simulation indicates that our model is accurate for not only the mean response time but also its quantiles, which are particularly relevant for deadline-driven applications. Further, we show that in the distributed setup, replication with canceling has the potential to reduce response times, even at relatively high utilization. We also find that it offers response times close to those of the centralized system, especially at medium-to-high request reliability. These findings support the use of replication with canceling as an effective mechanism for both fault- and delay-tolerance.
Idioma originalEnglish (US)
Título de la publicación alojadaICPE 2016 - Proceedings of the 7th ACM/SPEC International Conference on Performance Engineering
Páginas197-208
Número de páginas12
ISBN (versión digital)9781450340809
DOI
EstadoPublished - mar 12 2016
Publicado de forma externa
Evento7th ACM/SPEC International Conference on Performance Engineering, ICPE 2016 - Delft
Duración: mar 12 2016mar 16 2016

Conference

Conference7th ACM/SPEC International Conference on Performance Engineering, ICPE 2016
PaísNetherlands
CiudadDelft
Período3/12/163/16/16

Huella dactilar

Farms
Servers
Stochastic models
Scalability
Processing

All Science Journal Classification (ASJC) codes

  • Software
  • Computer Science Applications
  • Hardware and Architecture

Citar esto

Qiu, Z., Pérez, J. F., & Harrison, P. G. (2016). Tackling latency via replication in distributed systems. En ICPE 2016 - Proceedings of the 7th ACM/SPEC International Conference on Performance Engineering (pp. 197-208) https://doi.org/10.1145/2851553.2851562
Qiu, Zhan ; Pérez, Juan F. ; Harrison, Peter G. / Tackling latency via replication in distributed systems. ICPE 2016 - Proceedings of the 7th ACM/SPEC International Conference on Performance Engineering. 2016. pp. 197-208
@inproceedings{3b8093a700604f7a9627f09ab2f070ea,
title = "Tackling latency via replication in distributed systems",
abstract = "Consistently high reliability and low latency are twin requirements common to many forms of distributed processing; for example, server farms and mirrored storage access. To address them, we consider replication of requests with canceling - i.e. initiate multiple concurrent replicas of a request and use the first successful result returned, canceling all outstanding replicas. This scheme has been studied recently, but mostly for systems with a single central queue, while server farms exploit distributed resources for scalability and robustness. We develop an approximate stochastic model to determine the response-time distribution in a system with distributed queues, and compare its performance against its centralized counterpart. Validation against simulation indicates that our model is accurate for not only the mean response time but also its percentiles, which are particularly relevant for deadline-driven applications. Further, we show that in the distributed set-up, replication with canceling has the potential to reduce response times, even at relatively high utilization. We also find that it offers response times close to those of the centralized system, especially at medium-to-high request reliability. These findings support the use of replication with canceling as an effective mechanism for both fault- and delay-tolerance.",
author = "Zhan Qiu and P{\'e}rez, {Juan F.} and Harrison, {Peter G.}",
year = "2016",
month = "3",
day = "12",
doi = "10.1145/2851553.2851562",
language = "English (US)",
pages = "197--208",
booktitle = "ICPE 2016 - Proceedings of the 7th ACM/SPEC International Conference on Performance Engineering",

}

Qiu, Z, Pérez, JF & Harrison, PG 2016, Tackling latency via replication in distributed systems. En ICPE 2016 - Proceedings of the 7th ACM/SPEC International Conference on Performance Engineering. pp. 197-208, Delft, 3/12/16. https://doi.org/10.1145/2851553.2851562

Tackling latency via replication in distributed systems. / Qiu, Zhan; Pérez, Juan F.; Harrison, Peter G.

ICPE 2016 - Proceedings of the 7th ACM/SPEC International Conference on Performance Engineering. 2016. p. 197-208.

Resultado de la investigación: Contribución a libro /Tipo informe o reporteContribución en conferencia

TY - GEN

T1 - Tackling latency via replication in distributed systems

AU - Qiu, Zhan

AU - Pérez, Juan F.

AU - Harrison, Peter G.

PY - 2016/3/12

Y1 - 2016/3/12

N2 - Consistently high reliability and low latency are twin requirements common to many forms of distributed processing; for example, server farms and mirrored storage access. To address them, we consider replication of requests with canceling - i.e. initiate multiple concurrent replicas of a request and use the first successful result returned, canceling all outstanding replicas. This scheme has been studied recently, but mostly for systems with a single central queue, while server farms exploit distributed resources for scalability and robustness. We develop an approximate stochastic model to determine the response-time distribution in a system with distributed queues, and compare its performance against its centralized counterpart. Validation against simulation indicates that our model is accurate for not only the mean response time but also its percentiles, which are particularly relevant for deadline-driven applications. Further, we show that in the distributed set-up, replication with canceling has the potential to reduce response times, even at relatively high utilization. We also find that it offers response times close to those of the centralized system, especially at medium-to-high request reliability. These findings support the use of replication with canceling as an effective mechanism for both fault- and delay-tolerance.

AB - Consistently high reliability and low latency are twin requirements common to many forms of distributed processing; for example, server farms and mirrored storage access. To address them, we consider replication of requests with canceling - i.e. initiate multiple concurrent replicas of a request and use the first successful result returned, canceling all outstanding replicas. This scheme has been studied recently, but mostly for systems with a single central queue, while server farms exploit distributed resources for scalability and robustness. We develop an approximate stochastic model to determine the response-time distribution in a system with distributed queues, and compare its performance against its centralized counterpart. Validation against simulation indicates that our model is accurate for not only the mean response time but also its percentiles, which are particularly relevant for deadline-driven applications. Further, we show that in the distributed set-up, replication with canceling has the potential to reduce response times, even at relatively high utilization. We also find that it offers response times close to those of the centralized system, especially at medium-to-high request reliability. These findings support the use of replication with canceling as an effective mechanism for both fault- and delay-tolerance.

UR - http://www.scopus.com/inward/record.url?scp=85020211058&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85020211058&partnerID=8YFLogxK

U2 - 10.1145/2851553.2851562

DO - 10.1145/2851553.2851562

M3 - Conference contribution

SP - 197

EP - 208

BT - ICPE 2016 - Proceedings of the 7th ACM/SPEC International Conference on Performance Engineering

ER -

Qiu Z, Pérez JF, Harrison PG. Tackling latency via replication in distributed systems. En ICPE 2016 - Proceedings of the 7th ACM/SPEC International Conference on Performance Engineering. 2016. p. 197-208 https://doi.org/10.1145/2851553.2851562