Using PIAF on Challenge

Access- Accounts- Data Areas- Get your data there- Get back archived data- Hints on user functions- How to use data files on Challenge

Access to Challege

Challenge should be accessible from any platform running standard version of PAW, irrespective of the geographical location. At the PAW prompt, type:

   connect piaf d0cha.fnal.gov

In your PAW session:

   help piaf

should help with most of the problems.
In case of troubles with connecting, check with your local system administrator, if the node d0cha.fnal.gov is accesible from your location. Send mail (including error messages, if any) to PIAF consultant.

Accounts

To use PIAF on Challenge you need an account there. All accounts from the d0sgi clusters are valid on the Challenge too. If you need an account, send mail to the System Manager User name and password must be entered at the PIAF prompts. Upon a succesful completion the following message should appear:

  **** Welcome to the Piaf server @ d0cha ****
  **** Piaf slave server @ d0cha started ****
  **** Piaf slave server @ d0cha started ****
  **** Piaf slave server @ d0cha started ****
  **** Piaf slave server @ d0cha started ****

Data Areas

Disk space is subdivided into the 'data' area ( /data2, /input, /output ), the 'spool' area ( /spool[1-6] ) and 'project' area ( /proj[1-14] ).

/dataX ( X = [1-3] ) are used for the general purpose data sets, like 'global Ntuples' and/or for various subsets requested by physics groups.

/projX ( X = [1-14] ) are used for storage of physics groups' data sets. They are devided among physics groups. /proj[1-4] are assigned to QCD group, /proj[5-8] are assigned to WZ group, /proj[9-12] are assigned to TOP LJETS group, and /proj[13-14] are assigned to TOP ALLJETS group.
Here is the up-to-date proj disks usage information.

/spoolX ( X = [1-6] ) are used for a storage of data sets for current use by individual physicists. The data can be spooled there when needed and will be disk resident. The spool disks management system is now available. This system archives all the spool disks data to a robotic mass storage. Purge 'less often used' data sets in the spool area so that spool disks will maintain around 20% free space. The archived data sets can be retrieved to the spool area whenever needed. Your suggestions, comments are always welcome, please contact PIAF consultant.
Here is the up-to-date spool disks usage information.

Getting the data over to the Challenge

To copy over the Ntuple from your local machine to the challenge use the PUT command. The files should be stored in the spool disks.

   PUT filename spool1/filename

Notice:
1. If you do it from a remote location, and the data file is large, this step may take a while.
2. Do not use ftp or ZFTP for this purpose. PUT is better optimized, and will give you better performance.

To use this file for a subsequent analysis do:

   h/file LUN //piaf/spool1/filename

To make it all work you need a symbolic link from your piaf directory to the spool area.

Retrieval of archived data sets from the Mass Storage

A system for the retrieval of the archived data has been established. You can get the archive data from STK silo back to the spool disks on the piaf server by PAW command STAGE:

   stage filename

For detailed usage of this command, use "help stage" inside your PAW session or click here to see the help manual of this command.

Another way to retrieve your archived data sets can be done using command line in your interactive login session to the challenge. Do following:

   d0cha% setup d0paw
   d0cha% filestat                  
   d0cha% retrieve_file filename

Note: 
   1. "filestat" will produce a list of all your files available in the 
      CAPHSM robot and indicate which files are not in spool disks.
      "retrieve_file" will retrieve your file back from CAPHSM to the 
      original directory.  

   2. "filestat" can have filename as its argument to filter the resulting
      file list. Wildcards * and ? is allowed, but you have to protect them 
      from shell to avoid file name expansion. For example:
                 filestat "single_et*"
	         filestat \*run_1a\* 

   3. wildcards is now allowed when you use "retrieve_file". please use it
      carefully. If there are more than one file with the same name in 
      CAPHSM, they will all be retrieved and restored naturally to different 
      original directories unless you specify "-l" option which retrieves 
      only the latest version.

How to Access The Data on Challenge

Piaf is expecting the data to reside in a piaf subdirectory of your home directory, e.g. /home/john/piaf, this is actually a soft link pointing to directory /piafdisk/john/piaf, /piafdisk is dedicated to piaf server use and has quota 25MB/30MB set for each user. To access the large data files, which reside in general on different disks, make symbolics links to the required areas.
Here is what you need to do:

   1. log in to the Challenge
   2. cd piaf 
	( you should be in /piafdisk/your_name/piaf now )    
   3. ln -sf /spool1/you spool1
      ... 
      ln -sf /spool6/you spool6
   4. exit

After this is done, you can put the data there via

   PUT file spoolX/file      ( X is from 1 to 6 )

and use via

   h/file LUN //piaf/spoolX/file

   chain -my_chain
   chain my_chain //piaf/spoolX/file1
   chain my_chain //piaf/spoolX/file2
   chain my_chain //piaf/spoolX/file3

NOTICE: //piaf in the above example tells piaf that the data file residing on the Challenge should be analyzed. If, in the same PAW session you type:

   h/file LUN1 file
   chain my_chain file1 file2 file3

then the local files will be analyzed.

Some hints on the User Functions

Adding 77 to the selection function will invoke the native f77 compiler and dynamically link the routines to Piaf. This will give a performance gain of a factor 5 or more over Comis. Example:
```
          nt/loop 10 func.f77
```
Notice: your function should be in the file func.f

executing a selection of the form:

             SELECT.F.AND.MASS<1.5

          is in general much less efficient than:

             MASS<1.5.AND.SELECT.F

          especially if MASS is very often >= 1.5.

never use the following construct to create histograms in selection functions:

             ...
             IF (IDNEVT .EQ. 1) THEN
                CALL HBOOK1(...)
                ...
             ENDIF

          but:

             SAVE IFIRST
             DATA IFIRST /0/

             IF (IFIRST .EQ. 0) THEN
                IFIRST = 1
                CALL HBOOK1(...)
                ...
             ENDIF

always do nt/plot or nt/loop over a few events first to make sure your selection function is ok.

	     PAW> nt/loop //chain/10 select.f77() 1000

	In piaf, Stdout and Stderr produced by your selection functions are 
	redirected to some log files .piaf*.log in your ~/piaf directory first,
	then PIAF server send those log files back to the client. 

	If your selection function produces error messages for each event and 
	you are looping over many events, your log files will become so big 
	that it fills up the /home disk, your piaf processes will hung and 
	will not send your messages back to you and other people can't use 
	their piaf servers properly.

Do NOT connect to the piaf server more than once from different PAW session in the same time. Because each newly started piaf server will delete the old piaf log files thus may cause the running old server some problems.

Last Updated August 30, 1996