Improving QoS by employing multiple physical NoCs on MPSoCs

Detalhes bibliográficos
Autor(a) principal: Silva, Douglas Roberto Guarani da
Data de Publicação: 2016
Tipo de documento: Dissertação
Idioma: eng
Título da fonte: Biblioteca Digital de Teses e Dissertações da PUC_RS
Texto Completo: http://tede2.pucrs.br/tede2/handle/tede/6634
Resumo: Embedded systems adopt NoC-based MPSoCs since a large number of processing elements (PEs) enables the simultaneous execution of several applications, where some of these applications require real-time (RT) constraints. PEs communicate using messages in distributed memory MPSoCs. These messages can be classified as application messages, being the data generated by the applications, and management messages, used to ensure the correct operation of the platform. As the communication has a large impact on the application performance, an important concern in the design of MPSoCs is to improve the performance of the applications’ communication, particularly for RT applications. Two possible methods to optimize the communication performance includes: (i) prioritize the RT application messages over the messages generated by best-effort (BE) applications; (ii) isolate the application messages from the management messages, considering that complex MPSoCs require a large number of management services to meet the performance constraints. The NoC literature contains several works that differentiate traffic classes, proposing the isolation of these traffic classes by the use of multiple physical (MP) NoCs, reducing interferences among the flows belonging to different classes. The main goal of this work is to propose and to evaluate MP NoCs, with one network dedicated to the application messages and a second network for the management messages (MNoC). Based on the evaluation of the impact of the management traffic in the overall NoC communication, two different versions of M-NoCs are implemented and evaluated. Another important consideration for RT applications is to ensure that these applications meet their deadlines. The execution of these applications must have higher priority over the BE applications by dedicating more processing resources using a specialized RT scheduler. This work presents and evaluates an MPSoC platform capable of supporting both communication and computation QoS, being extensible for a large number of management services by to the use of MP NoCs. Results show that M-NoCs may be customized to have a small area overhead. The adoption of M-NoCs improves the communication performance, latency and jitter, even when the network used in the platform has QoS mechanisms (e.g. priority flows and circuit switching), by isolating the management traffic from the application traffic.
id P_RS_759b2ed8a63867e518608f64f1614561
oai_identifier_str oai:tede2.pucrs.br:tede/6634
network_acronym_str P_RS
network_name_str Biblioteca Digital de Teses e Dissertações da PUC_RS
repository_id_str
spelling Moraes, Fernando Gehm477.763.820-00http://buscatextual.cnpq.br/buscatextual/visualizacv.do?id=K4782943Z2024.925.540-58http://buscatextual.cnpq.br/buscatextual/visualizacv.do?id=K4334759J7Silva, Douglas Roberto Guarani da2016-05-03T16:38:28Z2016-03-03http://tede2.pucrs.br/tede2/handle/tede/6634Embedded systems adopt NoC-based MPSoCs since a large number of processing elements (PEs) enables the simultaneous execution of several applications, where some of these applications require real-time (RT) constraints. PEs communicate using messages in distributed memory MPSoCs. These messages can be classified as application messages, being the data generated by the applications, and management messages, used to ensure the correct operation of the platform. As the communication has a large impact on the application performance, an important concern in the design of MPSoCs is to improve the performance of the applications’ communication, particularly for RT applications. Two possible methods to optimize the communication performance includes: (i) prioritize the RT application messages over the messages generated by best-effort (BE) applications; (ii) isolate the application messages from the management messages, considering that complex MPSoCs require a large number of management services to meet the performance constraints. The NoC literature contains several works that differentiate traffic classes, proposing the isolation of these traffic classes by the use of multiple physical (MP) NoCs, reducing interferences among the flows belonging to different classes. The main goal of this work is to propose and to evaluate MP NoCs, with one network dedicated to the application messages and a second network for the management messages (MNoC). Based on the evaluation of the impact of the management traffic in the overall NoC communication, two different versions of M-NoCs are implemented and evaluated. Another important consideration for RT applications is to ensure that these applications meet their deadlines. The execution of these applications must have higher priority over the BE applications by dedicating more processing resources using a specialized RT scheduler. This work presents and evaluates an MPSoC platform capable of supporting both communication and computation QoS, being extensible for a large number of management services by to the use of MP NoCs. Results show that M-NoCs may be customized to have a small area overhead. The adoption of M-NoCs improves the communication performance, latency and jitter, even when the network used in the platform has QoS mechanisms (e.g. priority flows and circuit switching), by isolating the management traffic from the application traffic.Sistemas embarcados adotam MPSoCs baseados em NoCs visto que um número grande de elementos de processamento (PEs) permitem a execução simultânea de várias aplicações, onde algumas dessas aplicações necessitam de restrições de tempo real (RT). PEs comunicam-se utilizando troca de mensagens em MPSoCs com memória distribuída. Essas mensagens podem ser classificadas como mensagens de aplicação, sendo os dados gerados pelas aplicações, e mensagens de gerência, utilizadas para garantir a operação correta da plataforma. Visto que a comunicação possui um forte impacto no desempenho da aplicação, uma preocupação importante no projeto de MPSoCs é de melhorar o desempenho da comunicação das aplicações, especialmente para aplicações RT. Dois métodos possíveis para otimizar o desempenho de comunicação incluem: (i) priorizar as mensagens das aplicações de RT sobre as mensagens geradas por aplicações de melhor esforço (do inglês, best effort, BE); (ii) isolar as mensagens de aplicações das mensagens de gerência, considerando que MPSoCs complexos necessitam de um grande número de serviços de gerência para satisfazer os requisitos de desempenho. Na literatura sobre NoCs há vários trabalhos que diferenciam classes de tráfego, propondo o isolamento dessas classes de tráfego pela utilização de múltiplas NoCs físicas (do inglês, multiple physical NoCs, MP NoCs), reduzindo interferências entre fluxos pertencentes a classes diferentes. O principal objetivo deste trabalho é propor e avaliar MP NoCs, onde uma rede é dedicada para mensagens de aplicação e uma segunda rede é utilizada para mensagens de gerência (M-NoC). Baseado na avaliação do impacto do tráfego de gerência na comunicação da NoC, duas versões da M-NoC são implementadas e avaliadas. Outra consideração importante para aplicações RT é garantir que os deadlines dessas aplicações sejam satisfeitos. A execução dessas aplicações deve ser priorizada sobre as aplicações BE através do fornecimento de mais recursos de processamento utilizando um escalonador RT especializado. Esse trabalho apresenta e avalia uma plataforma MPSoC capaz de suportar QoS de comunicação e de computação, sendo extensível para um número grande de serviços de gerência pelo uso de MP NoCs. Resultados mostram que as M-NoCs podem ser personalizadas para terem um pequeno impacto de área. A utilização de M-NoCs melhora o desempenho de comunicação, latência e jitter, mesmo considerando que a plataforma já possui mecanismos de QoS (como fluxos prioritários e chaveamento de circuitos), pelo isolamento do tráfego de gerência do tráfego de aplicação.Submitted by Setor de Tratamento da Informação - BC/PUCRS (tede2@pucrs.br) on 2016-05-03T16:38:28Z No. of bitstreams: 1 DIS_DOUGLAS_ROBERTO_GUARANI_DA_SILVA_COMPLETO.pdf: 3535458 bytes, checksum: a287a27f46a74cc898a15dccbe61ef44 (MD5)Made available in DSpace on 2016-05-03T16:38:28Z (GMT). No. of bitstreams: 1 DIS_DOUGLAS_ROBERTO_GUARANI_DA_SILVA_COMPLETO.pdf: 3535458 bytes, checksum: a287a27f46a74cc898a15dccbe61ef44 (MD5) Previous issue date: 2016-03-03Conselho Nacional de Pesquisa e Desenvolvimento Científico e Tecnológico - CNPqapplication/pdfhttp://tede2.pucrs.br:80/tede2/retrieve/164671/DIS_DOUGLAS_ROBERTO_GUARANI_DA_SILVA_COMPLETO.pdf.jpgengPontifícia Universidade Católica do Rio Grande do SulPrograma de Pós-Graduação em Ciência da ComputaçãoPUCRSBrasilFaculdade de InformáticaMULTIPROCESSADORESPROCESSAMENTO EM TEMPO REALARQUITETURA DE COMPUTADORINFORMÁTICACIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAOImproving QoS by employing multiple physical NoCs on MPSoCsAprimorando QoS utilizando múltiplas NoCs físicas em MPSoCsinfo:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/masterThesis1974996533081274470600600600600-30085425104011491443671711205811204509-2555911436985713659info:eu-repo/semantics/openAccessreponame:Biblioteca Digital de Teses e Dissertações da PUC_RSinstname:Pontifícia Universidade Católica do Rio Grande do Sul (PUCRS)instacron:PUC_RSTHUMBNAILDIS_DOUGLAS_ROBERTO_GUARANI_DA_SILVA_COMPLETO.pdf.jpgDIS_DOUGLAS_ROBERTO_GUARANI_DA_SILVA_COMPLETO.pdf.jpgimage/jpeg3736http://tede2.pucrs.br/tede2/bitstream/tede/6634/4/DIS_DOUGLAS_ROBERTO_GUARANI_DA_SILVA_COMPLETO.pdf.jpg0185f2b8746fb3952fde74c79a13d9e8MD54TEXTDIS_DOUGLAS_ROBERTO_GUARANI_DA_SILVA_COMPLETO.pdf.txtDIS_DOUGLAS_ROBERTO_GUARANI_DA_SILVA_COMPLETO.pdf.txttext/plain156801http://tede2.pucrs.br/tede2/bitstream/tede/6634/3/DIS_DOUGLAS_ROBERTO_GUARANI_DA_SILVA_COMPLETO.pdf.txt1f9f046e3f32fa4e39772c4c41de31bfMD53ORIGINALDIS_DOUGLAS_ROBERTO_GUARANI_DA_SILVA_COMPLETO.pdfDIS_DOUGLAS_ROBERTO_GUARANI_DA_SILVA_COMPLETO.pdfapplication/pdf3535458http://tede2.pucrs.br/tede2/bitstream/tede/6634/2/DIS_DOUGLAS_ROBERTO_GUARANI_DA_SILVA_COMPLETO.pdfa287a27f46a74cc898a15dccbe61ef44MD52LICENSElicense.txtlicense.txttext/plain; charset=utf-8610http://tede2.pucrs.br/tede2/bitstream/tede/6634/1/license.txt5a9d6006225b368ef605ba16b4f6d1beMD51tede/66342016-05-03 20:00:45.889oai:tede2.pucrs.br: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Biblioteca Digital de Teses e Dissertaçõeshttp://tede2.pucrs.br/tede2/PRIhttps://tede2.pucrs.br/oai/requestbiblioteca.central@pucrs.br||opendoar:2016-05-03T23:00:45Biblioteca Digital de Teses e Dissertações da PUC_RS - Pontifícia Universidade Católica do Rio Grande do Sul (PUCRS)false
dc.title.por.fl_str_mv Improving QoS by employing multiple physical NoCs on MPSoCs
dc.title.alternative.por.fl_str_mv Aprimorando QoS utilizando múltiplas NoCs físicas em MPSoCs
title Improving QoS by employing multiple physical NoCs on MPSoCs
spellingShingle Improving QoS by employing multiple physical NoCs on MPSoCs
Silva, Douglas Roberto Guarani da
MULTIPROCESSADORES
PROCESSAMENTO EM TEMPO REAL
ARQUITETURA DE COMPUTADOR
INFORMÁTICA
CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
title_short Improving QoS by employing multiple physical NoCs on MPSoCs
title_full Improving QoS by employing multiple physical NoCs on MPSoCs
title_fullStr Improving QoS by employing multiple physical NoCs on MPSoCs
title_full_unstemmed Improving QoS by employing multiple physical NoCs on MPSoCs
title_sort Improving QoS by employing multiple physical NoCs on MPSoCs
author Silva, Douglas Roberto Guarani da
author_facet Silva, Douglas Roberto Guarani da
author_role author
dc.contributor.advisor1.fl_str_mv Moraes, Fernando Gehm
dc.contributor.advisor1ID.fl_str_mv 477.763.820-00
dc.contributor.advisor1Lattes.fl_str_mv http://buscatextual.cnpq.br/buscatextual/visualizacv.do?id=K4782943Z2
dc.contributor.authorID.fl_str_mv 024.925.540-58
dc.contributor.authorLattes.fl_str_mv http://buscatextual.cnpq.br/buscatextual/visualizacv.do?id=K4334759J7
dc.contributor.author.fl_str_mv Silva, Douglas Roberto Guarani da
contributor_str_mv Moraes, Fernando Gehm
dc.subject.por.fl_str_mv MULTIPROCESSADORES
PROCESSAMENTO EM TEMPO REAL
ARQUITETURA DE COMPUTADOR
INFORMÁTICA
topic MULTIPROCESSADORES
PROCESSAMENTO EM TEMPO REAL
ARQUITETURA DE COMPUTADOR
INFORMÁTICA
CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
dc.subject.cnpq.fl_str_mv CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
description Embedded systems adopt NoC-based MPSoCs since a large number of processing elements (PEs) enables the simultaneous execution of several applications, where some of these applications require real-time (RT) constraints. PEs communicate using messages in distributed memory MPSoCs. These messages can be classified as application messages, being the data generated by the applications, and management messages, used to ensure the correct operation of the platform. As the communication has a large impact on the application performance, an important concern in the design of MPSoCs is to improve the performance of the applications’ communication, particularly for RT applications. Two possible methods to optimize the communication performance includes: (i) prioritize the RT application messages over the messages generated by best-effort (BE) applications; (ii) isolate the application messages from the management messages, considering that complex MPSoCs require a large number of management services to meet the performance constraints. The NoC literature contains several works that differentiate traffic classes, proposing the isolation of these traffic classes by the use of multiple physical (MP) NoCs, reducing interferences among the flows belonging to different classes. The main goal of this work is to propose and to evaluate MP NoCs, with one network dedicated to the application messages and a second network for the management messages (MNoC). Based on the evaluation of the impact of the management traffic in the overall NoC communication, two different versions of M-NoCs are implemented and evaluated. Another important consideration for RT applications is to ensure that these applications meet their deadlines. The execution of these applications must have higher priority over the BE applications by dedicating more processing resources using a specialized RT scheduler. This work presents and evaluates an MPSoC platform capable of supporting both communication and computation QoS, being extensible for a large number of management services by to the use of MP NoCs. Results show that M-NoCs may be customized to have a small area overhead. The adoption of M-NoCs improves the communication performance, latency and jitter, even when the network used in the platform has QoS mechanisms (e.g. priority flows and circuit switching), by isolating the management traffic from the application traffic.
publishDate 2016
dc.date.accessioned.fl_str_mv 2016-05-03T16:38:28Z
dc.date.issued.fl_str_mv 2016-03-03
dc.type.status.fl_str_mv info:eu-repo/semantics/publishedVersion
dc.type.driver.fl_str_mv info:eu-repo/semantics/masterThesis
format masterThesis
status_str publishedVersion
dc.identifier.uri.fl_str_mv http://tede2.pucrs.br/tede2/handle/tede/6634
url http://tede2.pucrs.br/tede2/handle/tede/6634
dc.language.iso.fl_str_mv eng
language eng
dc.relation.program.fl_str_mv 1974996533081274470
dc.relation.confidence.fl_str_mv 600
600
600
600
dc.relation.department.fl_str_mv -3008542510401149144
dc.relation.cnpq.fl_str_mv 3671711205811204509
dc.relation.sponsorship.fl_str_mv -2555911436985713659
dc.rights.driver.fl_str_mv info:eu-repo/semantics/openAccess
eu_rights_str_mv openAccess
dc.format.none.fl_str_mv application/pdf
dc.publisher.none.fl_str_mv Pontifícia Universidade Católica do Rio Grande do Sul
dc.publisher.program.fl_str_mv Programa de Pós-Graduação em Ciência da Computação
dc.publisher.initials.fl_str_mv PUCRS
dc.publisher.country.fl_str_mv Brasil
dc.publisher.department.fl_str_mv Faculdade de Informática
publisher.none.fl_str_mv Pontifícia Universidade Católica do Rio Grande do Sul
dc.source.none.fl_str_mv reponame:Biblioteca Digital de Teses e Dissertações da PUC_RS
instname:Pontifícia Universidade Católica do Rio Grande do Sul (PUCRS)
instacron:PUC_RS
instname_str Pontifícia Universidade Católica do Rio Grande do Sul (PUCRS)
instacron_str PUC_RS
institution PUC_RS
reponame_str Biblioteca Digital de Teses e Dissertações da PUC_RS
collection Biblioteca Digital de Teses e Dissertações da PUC_RS
bitstream.url.fl_str_mv http://tede2.pucrs.br/tede2/bitstream/tede/6634/4/DIS_DOUGLAS_ROBERTO_GUARANI_DA_SILVA_COMPLETO.pdf.jpg
http://tede2.pucrs.br/tede2/bitstream/tede/6634/3/DIS_DOUGLAS_ROBERTO_GUARANI_DA_SILVA_COMPLETO.pdf.txt
http://tede2.pucrs.br/tede2/bitstream/tede/6634/2/DIS_DOUGLAS_ROBERTO_GUARANI_DA_SILVA_COMPLETO.pdf
http://tede2.pucrs.br/tede2/bitstream/tede/6634/1/license.txt
bitstream.checksum.fl_str_mv 0185f2b8746fb3952fde74c79a13d9e8
1f9f046e3f32fa4e39772c4c41de31bf
a287a27f46a74cc898a15dccbe61ef44
5a9d6006225b368ef605ba16b4f6d1be
bitstream.checksumAlgorithm.fl_str_mv MD5
MD5
MD5
MD5
repository.name.fl_str_mv Biblioteca Digital de Teses e Dissertações da PUC_RS - Pontifícia Universidade Católica do Rio Grande do Sul (PUCRS)
repository.mail.fl_str_mv biblioteca.central@pucrs.br||
_version_ 1799765318163234816