In one environment, SORT choose work datasets as its sort work files:
from the Joblog I could see
ICE054I 0 RECORDS - IN: 1568500, OUT: 1568500
ICE134I 0 NUMBER OF BYTES SORTED: 4100059000
ICE165I 0 TOTAL WORK DATA SET TRACKS ALLOCATED: 99225 , TRACKS USED: 75000
ICE180I 0 HIPERSPACE STORAGE USED = 0K BYTES
ICE188I 0 DATA SPACE STORAGE USED = 201884K BYTES
but in another environment, SORT choose memory to store its sort work files
the job log is
ICE054I 0 RECORDS - IN: 3361000, OUT: 3361000
ICE134I 0 NUMBER OF BYTES SORTED: 8785654000
ICE165I 0 TOTAL WORK DATA SET TRACKS ALLOCATED: 0 , TRACKS USED: 0
ICE199I 0 MEMORY OBJECT STORAGE USED = 8424M BYTES
Although the input files are not same, they are similar ... I cannot understand why SORT was choosing different type of storage ... can any factor impact it?
Joined: 15 Feb 2005 Posts: 7129 Location: San Jose, CA
DFSORT chooses dynamically among memory object sorting, hipersorting, dataspace sorting and work data sets based on many factors including what's happening in the system at the time a particular sort application is run. DFSORT may choose different options for the same job on different systems based on the available resources, or even on the same system at different times based on the current workload.
Note that in your first case, DFSORT chose a combination of dataspace sorting and work data sets, whereas in the second case DFSORT chose memory object sorting. Memory object sorting is only available with z/OS DFSORT V1R5 and above, not with DFSORT R14. You didn't show all of the messages, so I can't tell which release of DFSORT you were using for each run.
Joined: 15 Feb 2005 Posts: 7129 Location: San Jose, CA
I already did clarify when I said:
Quote:
DFSORT may choose different options for the same job on different systems based on the available resources, or even on the same system at different times based on the current workload.
If you had used:
//SORTDIAG DD DUMMY
in the job that didn't use Memory Object Sorting, DFSORT would have produced diagnostic messages which might have allowed me to tell you why it didn't use it. But without those diagnostic messages, I can't say.
If you can add the SORTDIAG DD and recreate the situation, send me all of the DFSORT messages offline (yaeger@us.ibm.com) and I'll take a look.
Quote:
In one environment, the SORT chose work data sets - see the ICE165I messages
Well, actually, for one run DFSORT chose hipersorting and work data sets, and for the other it chose memory object sorting.
Joined: 15 Feb 2005 Posts: 7129 Location: San Jose, CA
Again, R14 can choose hipersorting, dataspace sorting and/or work data sets, e.g. Hipersorting can be used alone or with work data sets. V1R5 adds memory object sorting alone or with work data sets.
You can turn off memory object sorting (V1R5 only) with the MOSIZE=0 option.
You can turn off data space sorting with the DSPSIZE=0 option.
You can turn off hipersorting with the HIPRMAX=0 option.