Computer cluster for quantum scientific calculations and molecular dynamics calculations (4 units) | TEGAKARI, an information dissemination media for research and development

We are considering introducing a computer cluster for quantum chemical calculations and molecular dynamics calculations using software such as GROMACS and LAMMPS.
I would like you to propose a configuration consisting of a system in which workstations (nodes) are parallelized, with 3 or more nodes and a total number of cores of 60 or more.

I would like to equip one of the nodes with a GPU for molecular dynamics calculations such as RTX A1.

Regarding software, we plan to use GAMESS, LAMMPS, GROMACS, RadonPy, and ERmod.
I would like the above software to be pre-installed and delivered in a state where jobs can be managed using PBS.
Furthermore, for LAMMPS and GROMACS, it is preferable that jobs can be executed using GPU.

The specific wishes for each node configuration are as follows.

[Management node]

・CPU: Operating clock 2.0GHz or more, total number of cores 12C/24T or more
・Memory: About 128GB
・Storage: 80TB or more
・OS: Rocky Linux
・Others: Rack mount body, display & keyboard

[GPU-equipped calculation node]

・CPU: 2 CPU configuration, operating clock 2.75GHz or more, total number of cores 56C/112T or more
・Memory: About 48GB
・Storage: 500GB or more
・OS: Rocky Linux
・GPU: Made by NVIDIA, CUDA cores 10000 or more, VRAM total 48GB or more

[Computation node]

・CPU: 2 CPU configuration, operating clock 2.75GHz or more, total number of cores 56C/112T or more
・Memory: About 256GB
・Storage: 500GB or more
・OS: Rocky Linux

Other

・Connection between nodes with 10GbE LAN
・I would like to fit each node into a server rack, so please make a proposal that includes the rack itself.
・Budget around 900 million yen
・Request for on-site installation work

Based on these requests, we proposed the following cluster configuration.

management node
CPU	Intel Xeon Silver 4310 (2.10GHz 12 cores)
memory	128GB REG ECC
storage	20TB HDD S-ATA × 5(RAID6)2
Video	on board (VGAx1)
network	on board (1000Base-T x2), LAN card (10GbE x2)
Housing + power supply	4U housing + 1200W
OS	Rocky linux
Other	1U Rackmount 750VA/680W UPS 10GbE 28-port network switch 23.8-inch wide FullHD LCD display

Compute node with GPU
CPU	AMD EPYC MILAN 7453 (2.75GHz 28 cores) x2
memory	256GB REG ECC
storage	500GB SSD S-ATA
Video	on board (VGAx1)
GPU	NVIDIA RTX A6000 48GB GDDR6
network	on board (1000Base-T x2), LAN card (10GbE x2)
Housing + power supply	4U housing + 1200W
OS	Rocky linux

Compute node
CPU	AMD EPYC MILAN 7453 (2.75GHz 28 cores) x2
memory	256GB REG ECC
storage	500GB SSD S-ATA
Video	on board (VGAx1)
network	on board (1000Base-T x2), LAN card (10GbE x2)
Housing + power supply	4U housing + 1200W
OS	Rocky linux

Other
ラック	42U server rack
Support	On-site installation

4computer cluster

A total of 2 configurations: management node + GPU-equipped compute node + non-GPU compute node x4
Configure the cluster.

The management node also serves as shared storage and is configured in RAID20 with 5 x 5TB HDD S-ATA.
As a computational cluster,1 cores per node It is a cluster system with x3 units and a total of 168 cores.
Additionally, the GPU-equipped compute nodes are equipped with NVIDIA RTX A6000 48GB GDDR6.

Each node, UPS, and network switch are housed in a 42U server rack, with on-site installation included.
Each node is connected via a 10GbE network switch.

Software setup with job management enabled

GAMESS, LAMMPS, GROMACS, RadonPy, and ERmod are preinstalled according to customer requests.
LAMMPS/GROMACS is set up so that it can perform calculations using GPU.
Additionally, PBS is installed on the management node and configured to distribute computing resources to each node.

The configuration of this case study is based on the conditions given by the customer.
We will flexibly propose machines according to your conditions, so please feel free to contact us even if you are considering different conditions than what is listed.

■ Keywords

・What is LAMMPS?
LAMMPS (Large-scale Atomic/Molecular Massively Parallel Simulator) is an open-source molecular dynamics calculation application developed by Sandia National Laboratories in the United States.It has a reputation for its wide range and flexibility of force fields that can be handled by parameter settings.

reference:LAMMPS Molecular Dynamics Simulator *Jumps to an external site

・What is GROMACS?
GROMACS (Groningen Machine for Chemical Simulations) is a molecular dynamics simulation package developed at the University of Groningen in the Netherlands.It is primarily designed for biochemical molecules such as proteins, lipids and diffusion.

reference:GROMACS *Jumps to an external site

・What is GAMESS?
GAMESS (General Atomic and Molecular Electronic Structure System) is computational chemistry software for electronic structural analysis of molecular structures and reactions. GAMESS can calculate physicochemical properties such as molecular electronic structures, reaction energies, and spectra using theories based on quantum mechanics.

reference:GAMES *Jumps to an external site

・What is RadonPy?
RadonPy is an open source Python library that can perform fully automated calculations of polymer physical properties using all-atom classical molecular dynamics methods.The entire process of inputting the chemical structure of a polymer's repeating unit, assigning a force field, generating an initial structure, handling errors, and evaluating physical properties through equilibrium/non-equilibrium MD calculations can be completely automated.

reference:RadonPy *Jumps to an external site

・What is ERmod?
ERmod (Energy Representation Module) is software for calculating the free energy of soft molecular collective systems.It is possible to determine solubility in a wide range of solvent types, including not only ordinary liquids but also supercritical fluids and ionic liquids.

reference:ERmod *Jumps to an external site

・What is PBS?
PBS (Portable Batch System) is software for job scheduling.Often used in conjunction with UNIX cluster environments, it allows computing tasks to be assigned to resources on the cluster.