Research Area:  Metaheuristic Computing
Many organizations, companies or universities have accumulated, over the years, a large number of computing resources grouped in Clusters. Cluster Federated Environments arise as a new architecture with the objective of joining all these resources, increasing the global computing capacity of the organization without making a great economic investment. However, the high number of machines and computing resources impels great energy consumption.
Due to the economic and sustainable connotations that this entails, recently a new line of investigation has focused on reducing energy consumption while maximizing the performance of the applications and the usage of the system. The scheduling in these environments, responsible for allocating the applications to the system resources, offers the possibility of obtaining great improvements, as managing the resources correctly can have a great impact on the system performance and energy efficiency. However, this process is very complex, since it belongs to the NP problem group.
This PhD studies the problem of scheduling large batch workloads extracted from diverse real traces. The proposed techniques consider the heterogeneity of the system resources as well as the ability to apply co-allocation in order to take advantage of the leftover resources across clusters. The proposals will use sophisticated multi-criteria tactics, based on Genetic Algorithms and Particle Swarm Optimization, focused on reducing both the execution time of the jobs and the energy consumption of the system.
Name of the Researcher:  Eloi Gabaldon Ponsa
Name of the Supervisor(s):  Josep Lluis Lerida
Year of Completion:  2017
University:  University Of Lleida
Thesis Link:   Home Page Url