Site Tools


anthill_sri005

FIXME

description

software version

sbatch example

...

performance tests

...

FIXME

Anthill NAMD 212
preparation

Below information allow you to prepare NAMD software on Anthill computing cluster.
download NAMD source

Source tarball NAMD_2.12_Source.tar.gz can be downloaded from : http://www.ks.uiuc.edu/Development/Download/download.cgi?PackageName=NAMD .
compile NAMD from source

By following bellow you will create NAMD in ~/anthill23_NAMD/source/NAMD_2.12_Source/Linux-x86_64-g++/ .

Login to anthill23.cent1.uw.edu.pl .

Create directory for NAMD :

[anteater@anthill23 ~]$ mkdir -p ~/anthill23_NAMD/source

Copy NAMD_2.12_Source.tar.gz to ~/anthill23_NAMD/source. ( Remember that pier23.cent1.uw.edu.pl and anthill23.cent1.uw.edu.pl share /home/users and /workspace ).

Start interactive session and go through compilation instructions :

[anteater@anthill23 ~]$ srun --pty -plong --ntasks=8 --mem=8G bash -l
anteater@ant007:~$ 

#untar source tarball
anteater@ant007:~$ cd ~/anthill23_NAMD/source
anteater@ant007:~/anthill23_NAMD/source$ tar xzf NAMD_2.12_Source.tar.gz
anteater@ant007:~/anthill23_NAMD/source$ cd NAMD_2.12_Source/
anteater@ant011:~/anthill23_NAMD/source/NAMD_2.12_Source$

#prepare charm ( this proces should end with [... charm++ built successfully. ] message.
anteater@ant011:~/anthill23_NAMD/source/NAMD_2.12_Source$ tar xf charm-6.7.1.tar
anteater@ant011:~/anthill23_NAMD/source/NAMD_2.12_Source$ cd charm-6.7.1/
anteater@ant011:~/anthill23_NAMD/source/NAMD_2.12_Source/charm-6.7.1$./build charm++ net-linux-x86_64 --no-build-shared --with-production

#test charm ( this proces should end with [... test 53: completed (0.13 sec), All tests completed, exiting ] message.
anteater@ant011:~/anthill23_NAMD/source/NAMD_2.12_Source/charm-6.7.1$ cd net-linux-x86_64/tests/charm++/megatest
anteater@ant011:~/anthill23_NAMD/source/NAMD_2.12_Source/charm-6.7.1/net-linux-x86_64/tests/charm++/megatest$ make pgm
anteater@ant011:~/anthill23_NAMD/source/NAMD_2.12_Source/charm-6.7.1/net-linux-x86_64/tests/charm++/megatest$ ./charmrun ++local +p4 ./pgm

#prepare fftw
anteater@ant007:~$ cd ~/anthill23_NAMD/source/NAMD_2.12_Source
anteater@ant007:~/anthill23_NAMD/source$ wget http://www.ks.uiuc.edu/Research/namd/libraries/fftw-linux-x86_64.tar.gz
anteater@ant007:~/anthill23_NAMD/source$ tar xzf fftw-linux-x86_64.tar.gz
anteater@ant007:~/anthill23_NAMD/source$ mv linux-x86_64 fftw

#prepare tcl
anteater@ant007:~$ cd ~/anthill23_NAMD/source/NAMD_2.12_Source
anteater@ant007:~/anthill23_NAMD/source$ wget http://www.ks.uiuc.edu/Research/namd/libraries/tcl8.5.9-linux-x86_64.tar.gz
anteater@ant007:~/anthill23_NAMD/source$ tar xzf tcl8.5.9-linux-x86_64.tar.gz
anteater@ant007:~/anthill23_NAMD/source$ mv tcl8.5.9-linux-x86_64 tcl

#compile NAMD
anteater@ant007:~$ cd ~/anthill23_NAMD/source/NAMD_2.12_Source
anteater@ant007:~/anthill23_NAMD/source$ ./config Linux-x86_64-g++ --charm-arch net-linux-x86_64
anteater@ant007:~/anthill23_NAMD/source$ cd Linux-x86_64-g++
anteater@ant007:~/anthill23_NAMD/source/Linux-x86_64-g++$ make

#test NAMD (src/alanin example is part of source tarball)
anteater@ant007:~$ cd ~/anthill23_NAMD/source/NAMD_2.12_Source/Linux-x86_64-g++
anteater@ant007:~/anthill23_NAMD/source/Linux-x86_64-g++$ ./charmrun ++local +p4 ./namd2 src/alanin

passwordless ssh login between nodes

charm used by namd requires passwordless login between allocated computing nodes durring NAMD calculations.

To enable passwordless login between nodes : logging (ssh) in to the Anthill cluster ssh your_login@anthill23.cent.uw.edu.pl and type the following commands ( ssh-keygen asks different questions, do not enter anything, just press ENTER ) :

ssh-keygen -t rsa
cat ~/.ssh/id_rsa.pub >> ~/.ssh/authorized_keys
chmod 600 ~/.ssh/authorized_keys

After performing the above commands one should check whether the desired effect has been achieved. Login to anthill cluster, and allocate two nodes in interactive session. Then login to the second allocated node. Login to second node should not require a password.

[anteater@anthill23 ~]$ srun -N2 --pty bash -l
[anteater@ant011 ~]$ echo $SLURM_NODELIST        
 ant[011-012]                             ## <--- !! your allocated node list probably will be different !!
[anteater@ant011 ~]$ ssh ant012
Welcome to Ubuntu 16.04.3 LTS (GNU/Linux 4.4.0-97-generic x86_64)

* Documentation:  https://help.ubuntu.com
* Management:     https://landscape.canonical.com
* Support:        https://ubuntu.com/advantage
[anteater@ant012 ~]$ exit
logout
Connection to ant012 closed.
[anteater@ant011 ~]$ exit
logout
[anteater@anthill23 ~]$

using NAMD program
computation batch file

Below file anthill23_namd.batch uses 2 computing nodes (ant011, ant012) to run f1atpase/f1atpase.namd (ATPase benchmark).

#!/bin/bash -l
#SBATCH --partition=all
#SBATCH --nodes=2
#SBATCH --nodelist=ant011,ant012
#SBATCH --ntasks=8
#SBATCH --mem 8192M

BINPATH="/home/users/${USER}/anthill23_NAMD/source/NAMD_2.12_Source/Linux-x86_64-g++"
CHARMRUN_BIN="${BINPATH}/charmrun"
NAMD2_BIN="${BINPATH}/namd2"
WORKDIR="/home/users/${USER}/anthill23_NAMD/testjobs"

# start your computation:
echo "computation started : $(date +%Y%m%d%H%M%S)"
cd ${WORKDIR} 
${CHARMRUN_BIN} ++verbose ++scalable-start ++mpiexec +p8 ${NAMD2_BIN} +setcpuaffinity +isomalloc_sync f1atpase/f1atpase.namd
echo "computation ended : $(date +%Y%m%d%H%M%S)"

NAMD performance on Anthill
example simulation files

Site http://www.ks.uiuc.edu/Research/namd/utilities/ in secition Example Simulations provides files for tests :

ApoA1 benchmark (92,224 atoms, periodic, PME)
ATPase benchmark (327,506 atoms, periodic, PME)
STMV (virus) benchmark (1,066,628 atoms, periodic, PME)

example simulation results

In this tests each node was occupied only by one job.
example simulation ApoA1 benchmark results
node name 	cores used 	average(s/step) 	result1 [s] 	result2 [s] 	result3 [s] 	average [s]
ant003 	4 	0.534 	303.58 	281.21 	280.94 	288.58
ant004 	4 	0.529 	280.22 	278.33 	278.21 	278.92
ant[003-004] 	8 	0.277 	166.07 	151.31 	150.69 	156.02
ant005 	8 	0.229 	125.50 	125.30 	126.19 	125.66
ant006 	8 	0.229 	126.21 	125.91 	126.26 	126.13
ant[005-006] 	16 	0.119 	70.72 	70.66 	86.77 	76.05
ant008 	28 	0.033 	23.50 	24.17 	23.77 	23.81
ant400 	16 	0.144 	83.93 	83.99 	84.79 	84.24
ant401 	16 	0.145 	84.00 	85.17 	84.78 	84.65
ant[400-401] 	32 	0.080 	62.23 	53.90 	54.23 	56.79
ant[400-402] 	48 	0.066 	46.64 	46.89 	47.01 	46.85
ant[400-403] 	64 	0.052 	50.17 	41.85 	41.90 	44.64
ant[400-404] 	80 	0.054 	38.58 	39.66 	38.69 	38.98
ant[400-405] 	96 	0.053 	38.16 	38.37 	37.80 	38.11
ant[400-406] 	112 	0.057 	37.38 	37.20 	38.58 	37.72
ant[400-407] 	128 	0.055 	37.00 	36.77 	36.91 	36.89
ant[400-408] 	144 	0.054 	37.39 	37.56 	38.11 	37.69
ant[400-409] 	160 	0.047 	32.94 	32.92 	33.76 	33.21
ant[400-410] 	176 	0.044 	31.47 	32.49 	32.73 	32.23
ant[400-411] 	192 	0.043 	31.57 	32.54 	32.65 	32.25
ant[300-311,400-411] 	384 	0.036 	34.31 	33.48 	31.37 	33.05
ant[110,200-211,300-311,400-411] 	576 	0.037 	35.90 	39.19 	40.37 	38.49
example simulation ATPase benchmark results
node name 	cores used 	average(s/step) 	result1 [s] 	result2 [s] 	result3 [s] 	average [s]
ant003 	4 	1.652 	875.36 	850.86 	860.43 	862.22
ant004 	4 	1.624 	840.07 	851.99 	841.40 	844.49
ant[003-004] 	8 	0.856 	481.53 	451.53 	450.43 	461.16
ant005 	8 	0.695 	382.86 	383.14 	380.12 	382.04
ant006 	8 	0.694 	380.98 	378.75 	378.57 	379.43
ant[005-006] 	16 	0.359 	205.20 	205.90 	219.30 	210.13
ant008 	28 	0.097 	77.58 	65.78 	65.60 	69.65
ant400 	16 	0.442 	250.13 	250.82 	250.50 	250.48
ant401 	16 	0.443 	247.72 	251.62 	248.26 	249.20
ant[400-401] 	32 	0.234 	179.52 	153.49 	154.16 	162.39
ant[400-402] 	48 	0.165 	115.49 	116.33 	129.08 	120.30
ant[400-403] 	64 	0.153 	108.87 	108.50 	110.08 	109.15
ant[400-404] 	80 	0.149 	101.84 	102.60 	114.56 	106.33
ant[400-405] 	96 	0.127 	95.04 	93.40 	93.53 	93.99
ant[400-406] 	112 	0.118 	87.95 	89.56 	92.37 	89.96
ant[400-407] 	128 	0.106 	84.18 	82.03 	81.11 	82.44
ant[400-408] 	144 	0.107 	80.67 	79.28 	79.94 	79.96
ant[400-409] 	160 	0.101 	75.88 	76.03 	76.00 	75.97
ant[400-410] 	176 	0.086 	72.00 	70.50 	72.34 	71.61
ant[400-411] 	192 	0.088 	71.51 	71.14 	74.37 	72.34
ant[300-311,400-411] 	384 	0.060 	99.98 	101.39 	114.63 	105.33
ant[110,200-211,300-311,400-411] 	576 	0.053 	106.08 	105.37 	103.73 	105.06
example simulation STMV (virus) benchmark results
node name 	cores used 	average(s/step) 	result1 [s] 	result2 [s] 	result3 [s] 	average [s]
ant003 	4 	5.206 	2707.34 	2677.83 	2700.85 	2695.34
ant004 	4 	5.165 	2661.45 	2662.53 	2656.00 	2659.99
ant[003-004] 	8 	2.681 	1405.75 	1406.41 	1408.86 	1407.01
ant005 	8 	2.296 	1210.01 	1211.15 	1207.16 	1209.44
ant006 	8 	2.292 	1207.05 	1207.00 	1206.27 	1206.77
ant[005-006] 	16 	1.179 	647.70 	651.09 	648.72 	649.17
ant007 	28 	- 	- 	- 	- 	-
ant008 	28 	0.319 	179.59 	180.58 	179.67 	179.95
ant00[7-8] 	56 	- 	- 	- 	- 	-
ant009 	16 	- 	- 	- 	- 	-
ant011 	4 	1.542 	791.81 	791.86 	792.39 	792.02
ant012 	4 	1.545 	793.29 	793.31 	792.05 	792.88
ant[011-012] 	8 	0.783 	409.48 	409.23 	409.61 	409.44
ant400 	16 	1.474 	799.03 	794.11 	798.76 	797.30
ant401 	16 	1.473 	800.22 	792.96 	795.09 	796.09
ant[400-401] 	32 	0.757 	465.90 	456.35 	452.27 	458.17
ant[400-402] 	48 	0.526 	343.55 	340.48 	353.90 	345.98
ant[400-403] 	64 	0.411 	284.37 	272.64 	272.48 	276.50
ant[400-404] 	80 	0.403 	259.74 	258.68 	271.77 	263.40
ant[400-405] 	96 	0.345 	230.11 	231.95 	232.16 	231.41
ant[400-406] 	112 	0.303 	211.68 	210.85 	225.07 	215.87
ant[400-407] 	128 	0.276 	195.10 	194.50 	194.65 	194.75
ant[400-408] 	144 	0.256 	185.53 	185.05 	185.59 	185.39
ant[400-409] 	160 	0.241 	173.16 	175.06 	172.77 	173.66
ant[400-410] 	176 	0.215 	164.48 	164.24 	164.68 	164.47
ant[400-411] 	192 	0.199 	156.21 	154.94 	155.36 	155.50
ant[300-311,400-411] 	384 	0.129 	160.57 	163.19 	163.13 	162.30
ant[200-211,300-311,400-411] 	576 	0.100 	156.82 	153.73 	- 	155.28
anthill_sri005.txt · Last modified: 2023/08/01 01:08 by 127.0.0.1