Nowadays, application scheduling is widely used as an
effective energy conservation method. The energy consumption
can be significantly reduced by consolidating applications on
as less servers as possible and making idle servers sleep or
power-off [3]. By optimizing the power consumption of the
network infrastructure, few of existing application scheduling
has been considered. The network infrastructure is built to
provide high bisection bandwidth for applications in typical
commodity large-scale systems. By sharing diverse applications and each utilizing a part of the system, every time many
of these networks remain unused