Loading...
Publication
Uma proposta de escalonamento de tarefas sensível ao contexto de aplicações móveis no paradigma fog computing
| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 6.27 MB | Adobe PDF |
Authors
Abstract(s)
Os pedidos de execução de aplicações na arquitetura cloud e no paradigma fog são
geralmente heterogéneos em termos de contextos ao nível dos dispositivos e das aplicações.
O escalonamento dos pedidos nessas arquiteturas é um problema de otimização com
múltiplas restrições. Apesar dos inúmeros esforços, o escalonamento de tarefas nessas
arquiteturas e paradigmas continua a apresentar alguns desafios aliciantes que nos levam a
questionar a forma como as tarefas são encaminhadas entre os diferentes dispositivos físicos,
nós da fog e cloud. A fog é definida como uma extensão da cloud, que disponibiliza serviços
de processamento, armazenamento e rede próximo da edge network, e devido à densidade e
heterogeneidade de dispositivos, o escalonamento é muito complexo e, na literatura,
encontramos ainda poucos estudos. Contrariamente, o escalonamento na cloud é
amplamente estudado. Diversos trabalhos de investigação abordam, no entanto, essa questão
na perspetiva de provedores de serviço ou otimizam os níveis da qualidade de serviço (QoS)
da aplicação. Ignoram, porém, informações contextuais ao nível do dispositivo e dos
utilizadores finais e as suas experiências de utilização (QoE).
Procurando trazer contributos inovadores nas áreas de escalonamento de tarefas e
computação distribuída, nesta tese, é proposta uma solução para o problema de
escalonamento de pedidos sensível ao contexto para o paradigma fog que minimiza os
tempos de execução da aplicação e maximiza as suas prioridades. Os diferentes parâmetros
de contexto são normalizados através da normalização Min-Max. A prioridade de cada
pedido é definida através da aplicação da técnica de análise Multiple Linear Regression
(MLR) e o seu escalonamento com vista a otimizar a QoE dos utilizadores, é feito recorrendo
a técnica de Otimização Multi-Objective Non-Linear Programming1
(MONLP). Os
resultados experimentais, encontrados a partir de simulações no kit de ferramentas iFogSim,
demonstram que a nossa proposta de escalonamento apresenta um melhor desempenho em
comparação com as propostas não sensível ao contexto (First Come First Served, Shortest
Job First e QoS-based), relativamente às métricas: percentagem de execução dos pedidos
com sucesso, tempo de espera e QoE dos utilizadores.
Application execution requests in cloud architectures and fog paradigm are generally heterogeneous in terms of device and application contexts, and the scheduling of requests in these architectures is an optimization problem with multiple constraints. On the other hand, despite the numerous efforts, task scheduling in these architectures continue to present some enticing challenges that lead us to question how tasks are routed between different physical devices, fog nodes and cloud. Fog is defined as an extension of the cloud, which provides processing, storage and network services near the edge network, and due to its density and heterogeneity of devices, the scheduling is very complex and in the literature, there are still few studies that have been conducted. Conversely, scheduling in the cloud has been widely studied. Nonetheless, many surveys address this issue from the perspective of service providers or optimize application quality of service (QoS) levels of the application. In addition, they ignore contextual information at the level of the device, end-users and their user experiences (QoE). Seeking to bring innovative contributions in the areas of task scheduling and distributed computing, in this thesis, we propose a solution to the problem of context-aware requisition scheduling for fog paradigm that minimizes application execution times (i.e. optimizes QoE) and maximizes its priorities. The different context parameters are normalized using MinMax normalization. The priority of each request is defined through the application of the Multiple Linear Regression analysis technique and the scheduling of the requests in order to optimize the users QoE, respecting the various constraints, is made using the multi-objective non-linear programming optimization technique. Our experimental results, obtained from simulation executions in the iFogSim toolkit, demonstrate that our scheduling proposal performs better than non-context-sensitive proposals (FCFS, SJF e QoS-based) in terms of metrics: success rate, waiting time and user QoE.
Application execution requests in cloud architectures and fog paradigm are generally heterogeneous in terms of device and application contexts, and the scheduling of requests in these architectures is an optimization problem with multiple constraints. On the other hand, despite the numerous efforts, task scheduling in these architectures continue to present some enticing challenges that lead us to question how tasks are routed between different physical devices, fog nodes and cloud. Fog is defined as an extension of the cloud, which provides processing, storage and network services near the edge network, and due to its density and heterogeneity of devices, the scheduling is very complex and in the literature, there are still few studies that have been conducted. Conversely, scheduling in the cloud has been widely studied. Nonetheless, many surveys address this issue from the perspective of service providers or optimize application quality of service (QoS) levels of the application. In addition, they ignore contextual information at the level of the device, end-users and their user experiences (QoE). Seeking to bring innovative contributions in the areas of task scheduling and distributed computing, in this thesis, we propose a solution to the problem of context-aware requisition scheduling for fog paradigm that minimizes application execution times (i.e. optimizes QoE) and maximizes its priorities. The different context parameters are normalized using MinMax normalization. The priority of each request is defined through the application of the Multiple Linear Regression analysis technique and the scheduling of the requests in order to optimize the users QoE, respecting the various constraints, is made using the multi-objective non-linear programming optimization technique. Our experimental results, obtained from simulation executions in the iFogSim toolkit, demonstrate that our scheduling proposal performs better than non-context-sensitive proposals (FCFS, SJF e QoS-based) in terms of metrics: success rate, waiting time and user QoE.
Description
Keywords
Aplicações móveis Qualidade de experiência Sensibilidade ao contexto Escalonamento de tarefas Cloud computing Fog computing Quality of experience Context awareness Task scheduling
Pedagogical Context
Citation
Barros, Celestino Lopes de - Uma proposta de escalonamento de tarefas sensível ao contexto de aplicações móveis no paradigma fog computing. [S.l.]: [s.n.], 2020. 189 p.