Workload placement on servers in clustered environments
has been traditionally driven by performance objectives.
All servers are intended to be fully operational in order to provide
the maximum available resources to users; the technique of
load balancing is used to spread the workload to as many of
them as possible. This mode of operation causes low utilization
of the cluster resources, and consequently consumes a high
amount of power. Research has pointed out that using techniques of load-screw scheduling and dynamically adjusting
idle servers to power-save mode can significantly reduce the
energy consumption