the relation between request arrival rate
and the percentage of servers operating in sleep mode has
been shown. The figure shows that, as the request arrival rate
increases, the number of the VMs also increases. However, if
we apply our proposed VSA, greater number of VMs can be
kept unused hence energy consumption reduces. Again, some
clusters can be kept in sleep mode at time of low request
arrival rate and as a result, the total VMs as well as physical
server at sleep state increases. Due to dynamic workload, VM
re-scheduling and considering bandwidth optimization at the
same time, better amount of energy can be saved by applying
VSA compared to EHS and ESI. As the rate continues to
increase, the difference in energy optimization degrades