République Tunisienne
Ministère de l'Enseignement Supérieur et RS Ministère des Technologies de la Communication et de l'Economie Numérique
Université de Carthage
La RechercheEspace membreCet espace est dédié aux étudiants, aux enseignants et au personnel administratif de l'école |
|
Evènements et manifestations 22/01/2018 Doctorate thesis defense of Meriam Essaies
Doctorate thesis defense on January 22th 2018 at 09H00 ,in Amphi Ibn Khaldoun, Sup’Com. Entitled :ON TASK SCHEDULING AND EXECUTION COST OPTIMIZATION IN CLOUD COMPUTING Presented by : Meriam Essaies Committee
AbstractNowadays cloud computing has become an emerging technology. Cloud computing has the potential to transform a large part of the IT industry, making software even more attractive as a service and shaping the way in which hardware is de-signed and purchased. Among the main challenges for the development of cloud technology in future, scheduling and re-source management problem stand out and attract our attention. Scheduling and resource management are complex tasks in cloud computing and are different from scheduling in grid computing. Scheduling in cloud is more critical as we pay as we go. Besides, in cloud computing we deal with virtual machines creation and release. In this work, we define scheduling and resource management techniques and we present its complexity. We mainly accomplish five research issues. First, we propose an algorithm for resource allocation that ensures the required QoS for the arriving tasks while minimizing the en-ergy cost. We propose to cluster the arriving task according to a cost function that reflects the sensibility of the task execu-tion to latency. After, the allocation of the needed resources to task clusters is based on a training model. Second, we focus on scheduling in Vehicular Cloud Computing and we propose a scheduling technique for critical application that aims to reduce latency for the most endangered vehicle’s applications. Besides, from cloud traces analysis, we address the re-source management problem and we propose a model drifted auto-scaling. It uses a time series resource utilization forecast and VM template to compute the needed resources to provision in order to meet the QoS. As for load balancing algorithm, we tried to highlight the problem of load computation uncertainty and its impact on cloud performance. To conclude, many techniques of scheduling have been proposed in cloud computing including load balancing, auto-scaling, energy saving, clustering, etc. We propose a fuzzy logic framework that, according to load state and task characteristics, will choose the adequate policy of scheduling to apply in the cloud system. KeywordsAuto-scaling, Clustering, Load Balancing, Migration, Resource management, Resource allocation, Scheduling, Vehicular cloud computing. |
|||||||||||||||||||||||||||||||||||||
