There is a feature in XFC where you can pull out statistics
 on how files were being cached on a particular device.  The
 information is interesting, but it's not in a form were it can
 be analyzed very well.  I've developed a procedure that
 converts it into a form where it can be loaded into a spread
 sheet.

$ SHOW MEMORY /CACHE = FILE = dkb300:*.* /OUT = xfc.log
$ EDI /TPU /NOSECT /NOINI /NODISP /COMMAND = xfc_cache_log_fix.tpu xfc.log
$ SORT /STAT /NODUP xfc.log xfc.log ! optional
$ COPY xfc.header, xfc.log xfc.csv

     The first step retrieves the data for a particular disk. 
 The second uses the TPU facility to manipulate the data.  (If
 symbol EDI is defined to be something else you'll have to
 override it to get 'raw' TPU).  The SORT command is optional. 
 The last command puts a header file on the data so you'll know
 what the fields are.

     This procedure works, but it leaves a few stray lines in
 the data.  These don't hurt, but I wanted to remove them. The
 data also contains the entire file specification as a single
 field, and I thought it would be easier to read if the device,
 directory, and file specification were separate fields.  I
 therefore wrote a DCL command procedure that processes the CSV
 data to clean it up a little, and to separate the various
 portions of the file specification.

     Bart.