EPOBF: Energy Efficient Allocation of Virtual Machines in High Performance Computing Cloud - Transactions on Large-Scale Data-and Knowledge-Centered Systems - page 81

baseline algorithm on problem of VM allocation for comparison in this paper.
The PABFD sorts all VMs by their current requested CPU utilization from
highest to lowest and then uses a Best-Fit Decreasing heuristic to select a phys-
ical machine for the first new virtual machine that has minimizing increasing
power on each placement of the VM to assign a new VM. The PABFD is pre-
sented in [ 6 ] and also denoted as MBFD in [ 5 ].

Fig. 2. The HPC simulated workload with starting times of 48,880 cloudlets

Table 2. Energy consumption (KWh)

VM Allocation
Heuristic

Num. of
Hosts

Num. of
VMs

Energy
(KWh)

Energy Saving (%)
+: better; -:worst

Shutdown
Hosts

PABFD 10000 48880 844.28 0% 21484

VBP Greedy L1 10000 48880 1072.86 -27% 23243

VBP Greedy L2 10000 48880 1072.86 -27% 23243

VBP Greedy L30 10000 48880 1072.86 -27% 23243

EPOBF-ST 10000 48880 664.16 21% 19025

EPOBF-FT 10000 48880 551.45 35% 18904

VBP greedy L1 , VBP greedy L2 ,and VBP greedy L30 : We implemented the
vector bin packing (VBP) heuristic that is presented as Norm-based Greedy
L1, Norm-based Greedy L2, and Norm-based Greedy L30 in [ 19 ]. All weights
of the Norm-based Greedy heuristics are proposed in and are implemented as
FFDAvgSum [ 19 ], which takes weights of resource dimensions that are exponen-
tial in average of the sum of demand resources.