Energy efficient web server clusters with stochastic QoS Control

Registro completo de metadados
MetadadosDescriçãoIdioma
Autor(es): dc.contributorLeite, Julius Cesar Barreto-
Autor(es): dc.contributorCPF:32908634122-
Autor(es): dc.contributorhttp://lattes.cnpq.br/4108398403192996-
Autor(es): dc.contributorMossé, Daniel-
Autor(es): dc.contributorCPF:74532167822-
Autor(es): dc.contributorPereira, Carlos Eduardo-
Autor(es): dc.contributorCPF:74680034522-
Autor(es): dc.contributorOliveira, Rômulo Silva de-
Autor(es): dc.contributorCPF:33090721322-
Autor(es): dc.contributorhttp://lattes.cnpq.br/5883242597607815-
Autor(es): dc.contributorMeira Junior, Wagner-
Autor(es): dc.contributorCPF:30219085522-
Autor(es): dc.contributorhttp://lattes.cnpq.br/9092587237114334-
Autor(es): dc.contributorLoques Filho, Orlando Gomes-
Autor(es): dc.contributorCPF:31345290822-
Autor(es): dc.contributorhttp://lattes.cnpq.br/9433123288261141-
Autor(es): dc.contributorBoeres, Maria Cristina Silva-
Autor(es): dc.contributorCPF:29609077322-
Autor(es): dc.contributorhttp://lattes.cnpq.br/0306766365983082-
Autor(es): dc.creatorBertini, Luciano-
Data de aceite: dc.date.accessioned2024-07-11T18:45:03Z-
Data de disponibilização: dc.date.available2024-07-11T18:45:03Z-
Data de envio: dc.date.issued2021-03-10-
Data de envio: dc.date.issued2010-03-25-
Data de envio: dc.date.issued2021-03-10-
Fonte completa do material: dc.identifierhttps://app.uff.br/riuff/handle/1/18752-
Fonte: dc.identifier.urihttp://educapes.capes.gov.br/handle/capes/776795-
Descrição: dc.descriptionFundação Carlos Chagas Filho de Amparo a Pesquisa do Estado do Rio de Janeiro-
Descrição: dc.descriptionIn this thesis we study the soft real-time web cluster architecture needed to support e-commerce and related applications, with the fundamental goal of optimizing the energy efficiency. The energy consumed by a system and its Quality of Service (QoS) are the two components of the tradeoff that rule the power optimization in such systems. In soft real-time systems like web servers, The QoS is usually defined as the fraction of requests that meet the deadlines. When this QoS is measured directly, regardless for whether the request missed the deadline by an small amount of time or by a large difference, result is always the same. For this reason, only counting the number of missed requests in a period does not allow an adequate observation of the real state of the system. We give theoretical propositions on how to control the QoS, not measuring the QoS directly, but based on the probability distribution of the tardiness in the completion time of requests . We call this QoS metric Tardiness Quantile Metric (TQM). The proposed method provides fine-grained control over the QoS so that we can make a closer examination of the relation between QoS and energy efficiency. To generalize the TQM idea, we propose the GTQM method that makes it possible to measure the QoS of the system without any assumption on the workload. By using an on line convergent sequential process defined from a Markov chain, we derive quantile estimations that do not depend on the shape of the workload probability distribution. To use GTQM, we need a controller that will keep the system's QoS as defined by the statistical inference. We describe a simplified way to implement performance control in a multi-tier computing system designed for e-commerce applications. We show that the simpler SISO (Single Input Single Output) controller, rather than a more complex distributed or centralized MIMO (Multiple Input Multiple Output) controller, works well, regardless of the presence of multiple cluster nodes and multiple executing time deadlines. Our feedback control lo-op acts on the speed of all server nodes capable of Dynamic Voltage Scaling (DVS), with QoS, measured by means of GTQM, being the reference setpoint. We use a SISO PIDF control loop implemented in the multi-tier cluster. Besides QoS control, we solve the dynamic configuration optimization problem in a web server cluster. We model the problem of selecting the servers that will be on and finding their speeds as mixed integer programming. For proof of concept, we implemented this dynamic configuration scheme and the GTQM QoS control in a web server cluster running Linux, and a layer of servers running a MySQL cluster configuration. The system has soft real-time requirements, in order to guarantee both energy-efficiency and good user experience. Our testbed is based on an industry standard, which defines a set of web interactions and database transactions with their deadlines, for generating real workload and benchmarking e-commerce applications.-
Formato: dc.formatapplication/pdf-
Idioma: dc.languageen-
Publicador: dc.publisherPrograma de Pós-Graduação em Computação-
Publicador: dc.publisherComputação-
Direitos: dc.rightsAcesso Aberto-
Direitos: dc.rightsCC-BY-SA-
Palavras-chave: dc.subjectCluster dynamic configuration-
Palavras-chave: dc.subjectDynamic voltage and frequency scaling-
Palavras-chave: dc.subjectEnergy efficiency-
Palavras-chave: dc.subjectEnergy efficient web server clusters-
Palavras-chave: dc.subjectPower aware real-time systems-
Palavras-chave: dc.subjectPower management-
Palavras-chave: dc.subjectQuality of service-
Palavras-chave: dc.subjectQoS control-
Palavras-chave: dc.subjectServer cluster-
Palavras-chave: dc.subjectSoft real-time systems-
Palavras-chave: dc.subjectWeb cluster-
Palavras-chave: dc.subjectComputação paralela e distribuída-
Palavras-chave: dc.subjectComércio eletrônico-
Palavras-chave: dc.subjectSistema em tempo real-
Palavras-chave: dc.subjectWeb - Aplicativo-
Palavras-chave: dc.subjectConsumo de energia-
Palavras-chave: dc.subjectCluster de processador-
Palavras-chave: dc.subjectAmbiente de configuração-
Palavras-chave: dc.subjectComputação eficiente em energia-
Palavras-chave: dc.subjectEscalonamento com otimização da potência-
Palavras-chave: dc.subjectCNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO::TEORIA DA COMPUTACAO::COMPUTABILIDADE E MODELOS DE COMPUTACAO-
Título: dc.titleEnergy efficient web server clusters with stochastic QoS Control-
Tipo de arquivo: dc.typeTese-
Aparece nas coleções:Repositório Institucional da Universidade Federal Fluminense - RiUFF

Não existem arquivos associados a este item.