Cloud vendors comparison

Date: 2016/09

Since each cloud provider uses a specific type of resources, performance per application may vary. For the sake of simplicity, a short VASP calculation utilizing 1 CPU with the following characteristics was used to quickly compare the performance of four cloud providers, AWS, Rackspace, SoftLayer and Microsoft Azure.

Model and Method

Plane-wave Pseudopotential Density Functional Theory formalism as implemented in Vienna Ab-initio Simulation Package (VASP) at version 5.3.5 with a corresponding set of atomic pseudo-potentials was employed in this run.

Inputs

INCAR

SYSTEM = Si
!!
NWRITE = 2
IALGO = 48
NELM = 13
ENMAX = 140 eV
IALGO = -1
NELMIN = 3
NELMDL = 7
NSIM = 4
LREAL = .TRUE.
BMIX = 2.5
ISYM = 0
EDIFF = 1E-4
LWAVE = .FALSE.
LCHARG = .FALSE.
!!
NSW = 0
POTIM = 5.00
TEBEG = 423
!!
ISMEAR = 1
SIGMA = 0.1
EMIN = -15
EMAX = 0

POSCAR

Silicon8
1.0
5.468728 0.000000 0.000000
0.000000 5.468728 0.000000
0.000000 0.000000 5.468728
Si
8
direct
0.250000 0.250000 0.250000 Si
0.750000 0.750000 0.250000 Si
0.500000 0.500000 0.000000 Si
0.000000 0.500000 0.500000 Si
0.250000 0.750000 0.750000 Si
0.000000 0.000000 0.000000 Si
0.750000 0.250000 0.750000 Si
0.500000 0.000000 0.500000 Si

KPOINTS

Automatic mesh
0
Gamma
  8  8  8
  0.  0.  0.

Results

Provider CPU Memory (GB) Disk (GB) Bandwidth (Gbps) Runtime (sec)
AWS 36 core, Intel Xeon E5-2666-v3, 2.90GHz 60 10 10 37.8
Azure 16 core, Intel Xeon E5-2673-v3, 2.40GHz 32 256 10 43.5
Rackspace 32 core, Intel Xeon E5-2680-v2, 2.80GHz 60 50 5 49
Softlayer 32 core, Intel Xeon E5-2650-v0, 2.00GHz 64 25 1 89.5

NOTE: no Infiband hardware was available on Azure at the time