DRM4G is an open platform, based on ?GridWay , to define, submit, and manage computational jobs. DRM4G is a ?Python (2.6+, 3.3+) implementation that provides a single point of control for computing resources without installing any intermediate middlewares. As a result, a user is able to run the same job on laptops, desktops, workstations, clusters, supercomputers, and any grid.
In order to install DRM4G, an installation script is provided. Type the command below on your shell terminal:
$ wget --no-check-certificate -O- https://meteo.unican.es/work/DRM4G/install.sh | bash ========================== DRM4G installation script ========================== --> Checking the last version of DRM4G ... --> DRM4G version selected: 2.4.1 --> Downloading drm4g-2.4.1.tar.gz ... --> Unpacking drm4g-2.4.1.tar.gz in directory /home/USER ... --> Installing DRM4G python requirements locally ... ==================================== Installation of DRM4G 2.4.1 is done! ==================================== In order to work with DRM4G you have to enable its environment with the command: . /home/USER/drm4g/bin/drm4g_init.sh You need to run the above command on every new shell you open before using DRM4G, but just once per session.
By default, it will install DRM4G on your current directory. But, you can download the installation script:
$ wget --no-check-certificate https://meteo.unican.es/work/DRM4G/install.sh
And run it manually:
$ bash ./install.sh [options]
The options available are:
If the the directory ~/.drm4g does not exist, drm4g will create one with a local configuration
[user@mycomputer~]$ . /home/user/drm4g/bin/drm4g_init.sh
[user@mycomputer~]$ drm4g start Creating a DRM4G local configuration in '/home/user/.drm4g' Creating '/home/user/.drm4g/var/acct' directory Coping from '/home/user/drm4g/etc' to '/home/user/.drm4g/etc' Starting DRM4G .... OK Starting ssh-agent ... OK
[user@mycomputer~]$ drm4g resource list RESOURCE STATE localmachine enabled [user@mycomputer~]$ drm4g host list HID ARCH JOBS(R/T) LRMS HOST 0 x86_64 0/0 fork localmachine [user@mycomputer~]$ drm4g host list 0 HID ARCH JOBS(R/T) LRMS HOST 0 x86_64 0/0 fork localmachine QUEUENAME JOBS(R/T) WALLT CPUT MAXR MAXQ default 0/0 0 0 1 1
[user@mycomputer~]$ echo "EXECUTABLE=/bin/date" > date.job
[user@mycomputer~]$ drm4g job submit date.job ID: 0
[user@mycomputer~]$ drm4g job list 0 JID DM EM START END EXEC XFER EXIT NAME HOST 0 pend ---- 19:39:09 --:--:-- 0:00:00 0:00:00 -- date.job --
0 pend ---- 19:39:09 --:--:-- 0:00:00 0:00:00 -- date.job -- 0 prol ---- 19:39:09 --:--:-- 0:00:00 0:00:00 -- date.job -- 0 wrap pend 19:39:09 --:--:-- 0:00:00 0:00:00 -- date.job localhost/fork 0 wrap actv 19:39:09 --:--:-- 0:00:05 0:00:00 -- date.job localhost/fork 0 epil ---- 19:39:09 --:--:-- 0:00:10 0:00:00 -- date.job localhost/fork 0 done ---- 19:39:09 19:39:27 0:00:10 0:00:01 0 date.job localhost/fork
[user@mycomputer~]$ cat stdout.0 Mon Jul 28 12:29:43 CEST 2014 [user@mycomputer~]$ cat stderr.0
In order to configure a TORQUE/PBS cluster accessed through ssh protocol, you can follow the next steps:
[user@mycomputer~]$ ssh-keygen -t rsa -b 2048 -f $HOME/.ssh/meteo_rsa -N ""
[user@mycomputer~]$ ssh-copy-id -i $HOME/.ssh/meteo_rsa.pub user@ui.macc.unican.es
[user@mycomputer~]$ drm4g resource edit [meteo] enable = true communicator = ssh username = user frontend = ui.macc.unican.es private_key = ~/.ssh/meteo_rsa lrms = pbs queue = qrid max_jobs_running = 1 max_jobs_in_queue = 2
[user@mycomputer~]$ drm4g resource list RESOURCE STATE meteo enabled [user@mycomputer~]$ drm4g host list HID ARCH JOBS(R/T) LRMS HOST 0 x86_64 0/0 pbs meteo
That's it! Now, you can summit jobs to meteo.
In this section it will be described how to take advantage of DRM4G to calculate the number Pi. To do that, three types of jobs single, array and mpi will be used.
#include <stdio.h> #include <string.h> #include <stdlib.h> int main (int argc, char** args) { int task_id; int total_tasks; long long int n; long long int i; double l_sum, x, h; task_id = atoi(args[1]); total_tasks = atoi(args[2]); n = atoll(args[3]); fprintf(stderr, "task_id=%d total_tasks=%d n=%lld\n", task_id, total_tasks, n); h = 1.0/n; l_sum = 0.0; for (i = task_id; i < n; i += total_tasks) { x = (i + 0.5)*h; l_sum += 4.0/(1.0 + x*x); } l_sum *= h; printf("%0.12g\n", l_sum); return 0; }
#include <stdio.h> #include <string.h> int main (int argc, char** args) { int task_id; int total_tasks; long long int n; long long int i; double l_sum, x, h; task_id = atoi(args[1]); total_tasks = atoi(args[2]); n = atoll(args[3]); fprintf(stderr, "task_id=%d total_tasks=%d n=%lld\n", task_id, total_tasks, n); h = 1.0/n; l_sum = 0.0; for (i = task_id; i < n; i += total_tasks) { x = (i + 0.5)*h; l_sum += 4.0/(1.0 + x*x); } l_sum *= h; printf("%0.12g\n", l_sum); return 0; }
EXECUTABLE = pi
ARGUMENTS = ${TASK_ID} ${TOTAL_TASKS} 100000
STDOUT_FILE = stdout_file.${TASK_ID}
STDERR_FILE = stderr_file.${TASK_ID}
#include "mpi.h" #include <stdio.h> #include <math.h> int main( int argc, char *argv[]) { int done = 0, n, myid, numprocs, i; double PI25DT = 3.141592653589793238462643; double mypi, pi, h, sum, x; double startwtime = 0.0, endwtime; int namelen; char processor_name[MPI_MAX_PROCESSOR_NAME]; MPI_Init(&argc,&argv); MPI_Comm_size(MPI_COMM_WORLD,&numprocs); MPI_Comm_rank(MPI_COMM_WORLD,&myid); MPI_Get_processor_name(processor_name,&namelen); printf("Process %d on %s\n", myid, processor_name); n = 100000000; startwtime = MPI_Wtime(); h = 1.0 / (double) n; sum = 0.0; for (i = myid + 1; i <= n; i += numprocs) { x = h * ((double)i - 0.5); sum += 4.0 / (1.0 + x*x); } mypi = h * sum; MPI_Reduce(&mypi, &pi, 1, MPI_DOUBLE, MPI_SUM, 0, MPI_COMM_WORLD); if (myid == 0) { printf("pi is approximately %.16f, Error is %.16f\n", pi, fabs(pi - PI25DT)); endwtime = MPI_Wtime(); printf("wall clock time = %f\n", endwtime-startwtime); } MPI_Finalize(); return 0; }
Job template :
EXECUTABLE = mpi
STDOUT_FILE = stdout.${JOB_ID}
STDERR_FILE = stderr.${JOB_ID}
NP = 2