DMRG

Summary

This guide illustrates how to run the DMRG application using IGP (Italian Grid Portal).

Prerequisites

It is mandatory a valid X509 certificate and VO GRIDIT membership. We assume that you are familiar with concepts like certificate proxies, job submission, storage elements. If this is not the case, better read this introduction. Other than that, you don't need to access an User Interface, since your grid interactions will be performed through IGP. Of course you will need to be registered to it.

Input File Upload

The first thing you need to do is the upload of your input files. They need to be grouped in a single tgz archive, where each file is named 1.inp, 2.inp, etc... You can choose freely a name for the tgz archive, that anyway will be set later. Once you created the archive on your machine, login to IGP and move to the Storage section of IGP, selecting the directory /grid/gridit/DMRG. Click on Upload, and select the archive just created. This step will upload the archive to a Grid Storage Element. You can leave the choice to IGP, or select one of your choice. Once done, press start to begin the transfer.

You might want to use your own executable file, the one carried out with IGP works only on Scientific Linux 6. Once completed the upload, the DMRG folder should look like here

Workflow

Import

From the top bar menu, select Workflow => Import. Click on Show list and select the workflow DMRG-parametric, and then import it.

Job Monitor

Now move to Workflow => Job Monitor, and choose Configure for the job just imported.

Configuration

Using the Generator option, you can set some parameters representing the application runs. You can set parameters range (-p ), starting offset (-s) and increment (-t). For instance, to program a parameter sweep from 1 to 10, with a increment of 1, you should set -p 10 -s 1 -t 1. With such settings, 10 jobs will be executed, each of them expecting as input 1.inp 2.inp......10.inp, that of course will be searched in the archive uploaded in the former step. By default all options are equal to one, so -p 10 -s 1 -t 1 and -p 10 would have identical effects, while leaving it empty will produce the execution of a single job, expecting input 1.inp.

Once parameters are set, save them using the green checkbox (V) and then close the window, which takes you back to the workflow step. Now clicking on the Parametric box, you can select the input files, and choose the output folder.

For instance, in this case Parameter contains three strings, separated with a white space : dmrg-7.8-SL6.tgz Heisenberg.tgz test2. The first is the name of the tgz containing the executable, the second is the name of the tgz containing the input files (1.inp, 2.inp, etc...) and the third is the name of the folder where the output will be saved. All these filenames refer to the path (on the Grid) /grid/gridit/DMRG (/output/ for the output).

Once that Parametric is edited, save again with the green checkbox and close the window to return back on the workflow screen. Save the workflow created clicking on the floppy icon and go back to the job monitor with the left arrow on top left.

Submit

From Job Monitor, from the entry related to drmg-parametric click on Submit. Now with Info you can check the job status. First being executed is the generator, and then the parametric (4 in this example)

with View Running

Output

If jobs have been executed correctly, they will enter Finished status. Now go to the Storage section, and in the folder /grid/gridit/DMRG/output you should see a subfolder with the name you have set as third parameter in the Parametric configuration (test2 in that case).

the folder contains the output and error streams. The tgz naming follows the convention parameter-index-seconds-since-epoch. If one or more jobs have the same output folder, there will be more files related to the same parameter, but you can still distinguish them since they'll have a different epoch time.

-- EmidioGiorgio - 2013-11-07