Résumé:
Cloud computing is getting more popular day by day. Nowadays, we have an energy
shortage worldwide. Thus, more researches concern about finding ways to improve the
energy efficiency and sustainability of cloud computing. This thesis presents a performance
analysis of the cloud computing system based on the energy-efficient task scheduling
strategy. This strategy consists of waking up threshold policy and a sleep delay timer.
To model the system in an energy-efficient task scheduling strategy, we use a
vacation queue system with an N-policy threshold. in this work, the system considered as
a physical machine with a limited buffer, this physical machine contains a set of virtual
machines that represent the servers in our modelization, as well we constructed a
continuous-time Markov chain and an infinitesimal generator that allows us to calculate
stationary distribution vector. After that, we analyze the steady-state system to calculate
the system's different performance measures like energy consumption and the delay of the
system.
We tested our system with different scenarios and with different parameter values
to discover the proposed strategy performance and validate the system model according to
the performance measure.
The experimental study has proven that the proposed solution improves energy
efficiency and latency without affecting availability in the system
Key words: Cloud Computing, energy efficiency, task scheduling, waking up threshold
sleep delay, queue system, Markov chain
