[Swift-commit] r3124 - SwiftApps/SEE/trunk

[noreply at svn.ci.uchicago.edu]

Author: aespinosa
Date: 2009-10-09 15:23:16 -0500 (Fri, 09 Oct 2009)
New Revision: 3124

Modified:
   SwiftApps/SEE/trunk/README
   SwiftApps/SEE/trunk/ampl.swift
Log:
Documentation on how to run the workflow

Modified: SwiftApps/SEE/trunk/README
===================================================================
--- SwiftApps/SEE/trunk/README	2009-10-08 17:08:46 UTC (rev 3123)
+++ SwiftApps/SEE/trunk/README	2009-10-09 20:23:16 UTC (rev 3124)
@@ -1,36 +1,36 @@
-In order to run ampl.swift, there are two lines of code we need to change.
-1) Open template with Emacs, VI, or any other editor
-The first line is option ampl_include "/work/00946/zzhang/SEE/static/instances/test/@INSTANCE_ID@\"
- Chnage the "/work/00946/zzhang" to your working dir.
+AMPL Workflow
+Prepared by : Allan Espinosa (aespinosa at cs.uchicago.edu)
 
-2) Open tc.data with Emacs, VI, or any other editor
-Change the 3rd coloumn last two entries
- localhost       run_ampl    /work/00946/zzhang/SEE/static/run_ampl      INSTALLED       INTEL32::LINUX  null
- tgtacc       run_ampl    /work/00946/zzhang/SEE/static/run_ampl        INSTALLED       INTEL32::LINUX  null
-to 
- localhost       run_ampl    /YOUR_WORK_DIR/SEE/static/run_ampl      INSTALLED       INTEL32::LINUX  null
- tgtacc       run_ampl    /YOUR_WORK_DIR/SEE/static/run_ampl        INSTALLED       INTEL32::LINUX  null
+This document demonstrates how to run and deploy an AMPL workflow on the
+Teraport compute resource. 
 
-3) Open tgranger.xml with Emacs, VI, or any other editor
-Change the following 3 entries accordingly
-    <profile namespace="globus" key="project">TG-CCR080022N</profile>
-    <workdirectory >/work/00946/zzhang/work</workdirectory>
-    <profile namespace="env" key="SWIFT_JOBDIR_PATH">/tmp/zzhang/jobdir</profile>
+1.  Prepare a site to run the workflow.  Copy the site_pack directory in the
+site.  For this example, let us say we deployed it in
+/home/USER/ampl-app/site_pack
 
-project is the project ID on ranger
-workdirectory is the shared work place for swift jobs
-SWIFT_JOB_DIR: if you don't set this, all swift job dirs will be created in workdirectory;
-               if you set it to /tmp/$USERANME/jobs, swift job dirs will be created on the ramdisk of the compute nodes.
-               By $USERNAME, I mean the real user name of you on ranger.
+2.  Update tc.data entries on TERAPORT about the run_ampl program 
+TERAPORT       run_ampl  /home/USER/ampl-app/site_pack/run_ampl        INSTALLED       INTEL32::LINUX  null
 
-4) All instances, 100 or 2000 test cases should be in instances/test/.
-Now, we have the 2000 test case in instances/test/
-It looks like run1, run2, run3, ..., run2000.
+3.  Update the sites.xml file as necessary.  This repository contains a sample
+file for deployment on TERAPORT
 
-To test ampl.swift local
-run "swift -tc.file ./tc.data ampl.swift"
+4.  Stage your data manually to TERAPORT.  A dataset contains directories named
+run1, run2, run3, etc.  For this example, we copy the dataset in
+/home/USER/ampl-data containing 3 run directories.
 
-To test ampl.swift on ranger
-run "swift -sites.file ./tgranger-sge-gram2.xml -tc.file ./tc.data ampl.swift"
+$ls /home/USER/ampl-data
+run1, run2, run3
 
-Good Luck
+5.  Update ampl.swift line 47 to reflect the size of the dataset.  In this case
+we set
+
+int runs[]=[1:3];
+
+6.  Update line 45 of ampl.swift to reflect the location of your dataset.  In
+this example se set
+
+string data_dir = "/home/USER/ampl-data"; 
+
+7.  Run the workflow by invoking swift.  You can see example runs in runampl.sh
+
+swift -tc.file tc.data -sites.file sites.xml ampl.swift

Modified: SwiftApps/SEE/trunk/ampl.swift
===================================================================
--- SwiftApps/SEE/trunk/ampl.swift	2009-10-08 17:08:46 UTC (rev 3123)
+++ SwiftApps/SEE/trunk/ampl.swift	2009-10-09 20:23:16 UTC (rev 3124)
@@ -41,8 +41,10 @@
 }
 
 AmplCmd const_cmd <ext;exec="./cmd_mapper.sh">;
+/* Update this to reflect the location of your dataset" */
 string data_dir = "/panfs/panasas/CMS/data/SEE/rdraws/all";
 
+/* Update this to reflect the length of your dataset */
 int runs[]=[1:5000];
 foreach i in runs {