Empirical evaluation of vector bin packing algorithms for energy efficient data centers

As the proliferation of cloud data centers has been progressing in recent years, the power consumed in cloud enterprises has increased also. Strategies to reduce the power consumption of a data center can be modelled into variants of vector bin packing algorithms. The application of virtualization technology allows a large number of virtual machines (VMs), treated as items, to be concurrently hosted into relatively fewer number of Physical Machines (PMs), treated as bins, as long as the sum capacity of the co-located VMs does not exceed the capacity of a hosting PM. This paper addresses the problem of which vector bin packing strategy to use in attempting to first consolidate VMs in a data center, the aim of which is to reduce power consumption by powering off unused PMs, then going forward, which strategy to adopt when packing new VM requests to PMs as optimally as possible.