Mémoires de Fin d’Etudes
Etablissement
Université de Béjaia - Abderrahmane Mira
Affiliation
Département d’Informatique
Auteur
BARKAT, Amine
Directeur de thèse
KECHADI Mohand Tahar (Professeur)
Filière
Informatique
Diplôme
Doctorat
Titre
Efficient Economical Model and Energy Consumption in the Clouds
Mots clés
Cloud Computing, Cluster Computing, Energy Consumption
Résumé
Unlike typical datacenters, where FLOPS is the only goal for optimization, cloud computing are formalized differently, under both technical and economic constraints. New terminologies have been introduced from economics perspective, such as FLOPS per-dollar per-watt. A concern to be addressed while targeting specific applications is that the system performance does not scale at the same rate as the price. Switching from a single core server to a dual core server might only increase the performance by 20% while doubling the price! There is also much work to do to have a detailed understanding of how energy is consumed in cloud contexts. A detailed breakdown of energy utilisation in typical datacenters can enable a cloud provider to include energy costs in there models that ultimately lead to more efficient applications. To address specifications, economics, and energy issues, this thesis will undertake extensive benchmarking close to the nature of a wide range of applications (including in solving linear equation, Mont Carlo simulation, embarrassingly parallel benchmarks, etc.) in addition to profiling tools, dynamic load management, and so on, in order to assess the capability of any deployment on target cloud infrastructures. Challenges include: how do this applications vary in terms of scaling up and down, and the energy consumption, for a given cloud configurations? Does it make sense to increase the resources? It is interesting to develop models of application consumption, leading to a classification, from the above and use them to determine total costs for different loads on different cloud configurations. This thesis will achieve this by an adaptive approach through the ability to leverage different underlying specifications into a unified deployment framework, on which the user can make informed decisions, and have a range of mapping choices in terms efficient economics and energy consumption.
Statut
Vérifié