• Resource allocation
• Genetic algorithm
• Data center
• SDN

In the recent years, data centers changed the way to provide hardware and software resources for high performance, scientific and business computing: with this facility is possible to reach good performances or virtually unlimited computational power without buying the whole needed infrastructure.
Data center owner instead has to deal with different problematics respect to end user like for example hardware maintenance and redundancy (because with a large number of devices fault probability of some of them increases), energy and power consumption needed to keep turned on the whole infrastructure, hot air dissipation and cooling for servers and switches and also other internal organization problems regarding for example the proper design of a network topology without bottlenecks providing the best quality of service as possible.
In this thesis is introduced a task allocation algorithm for data centers aiming to find a reasonable trade off between task’s completion time and devices power consumption.
This algorithm is designed using Genetic heuristics that allow both to explore solutions space and to search for the optimal solution in an efficient manner, and it is implemented on a dedicated framework for multi-objective Genetic algorithms, called jMetal.
Network flows are allocated and managed with the help of Software Defined Networking (SDN) architecture. SDN decouples control plane from data plane in switches; SDN control plane is centralized and every switch receives the proper forwarding rules according to the controller network view. Through this approach is possible to allocate perfectly connections in the network avoiding congestions and bottlenecks, as first step to realize energy saving also in the networking part.