CSV File format

The summary data spread-sheet file contains a comma separated list of
most of the values from the summary information.  These values come
from the files tstat, users and patterns. They are as follows:


 The first 8 values come from the name of the file, and are: 
+ the actual file name e.g. host_directory/full_name.gz
+ the host name
+ the domain name
+ the year
+ the month
+ the day
+ the hour
+ the minutes 
+ the number of decreasing time stamp warnings
+ the number of bogus byte warnings

Information from tstat  
======================
                
+ the word tstat
+ host IP address
+ kernel trace, agent and collector versions
+ boot time for the host 
+ trace  file's start time
+ trace file's stop time
+ total bytes in the trace file
+ number of raw records --  (split records count as 2 records here)
+ the number of operations per second
+ number of records -- (split records count as 1 records here)

  Information from the table of tstat 
  =================================== 

  The rows of this file are displayed either with there values from
  columns 2,4,5,6,7,8 the table presented from tstat, or for those
  system calls that don't directly involve a file-system (specifically,
  SETTIMEOFDAY EXIT FORK SETREUID SETTIMEOFDAY UTIMES) simply with
  their values from column 2.  These are specifically the
  number of records of the specified operation cumulatively, for
  failures, for ufs file systems, for afs file systems, for Coda file
  systems and for nfs file systems.


Information from the patterns file: (columns GM --HM)
==================================
 Patterns reads a trace and prints a summary of file access patterns
based on close records occurring in the trace file. The summary
includes the number of read-only, write-only, and read-write accesses
to files, as well as bytes transferred for each access type. In
addition, each of these access type is further broken down into
whole-file tranfer, other sequential access, and random access. Only
close records with a reference count of 1 are used in this summary,
because the statistics reported on close are cumulative.

 This information takes the format of 3x8 items.  For each access type,
Read-only, Write-only and Read-write  the following 8 items are listed.
+ Number of  read-only accesses 
+ number of bytes accessed read-only 
+ number of whole file, read-only accesses
+ number of bytes for the above accesses
+ number of read-only accesses that were sequential but not whole file
+ number of bytes for the above accesses
+ number of  read-only accesses that were random
+ number of bytes for the above accesses

This then repeats for Write-only and then Read-write.  Finally,
section then adds to items, the total number of accesses considered by
this analysis program and the total number of bytes considered.

Information from the users file:  
================================ 
 The next 10x3 fields indicate values from the users program.  The 10
most active user entries are added here, since the number of users is
variable.  If there are less than 10 total users, no-user is used to
indicate that this field is empty.  Additionally, the value of unknown
is provided by the users program for events in the traces that are
un-able to determine the user. For each entry the following items are
provided:

+ user the user-ID, unknown or no-user
+ the number of processes for that user
+ the number of records generated by that user

 This section ends with the uid  of the primary user (the most active
user with a uid > 99).  If no such user exists the field contains no-user.