MVSFORUMS.com

Fin · Beginner Joined: 19 Jan 2004 Posts: 27 Topics: 8

I have been asked to consider changing how data is input to a program:
currently it is in a VSAM file (about 6000 records) and I've been asked about putting this data in a regular COBOL internal table -

the advantages I am aware of would be:
less input / output leading to better performance
less maintenance

the downside would be an un-indexed table that might be more difficult to search (with only 6000 records this should not be an issue?)

Are the points above correct / valid?
Is there anything else I should consider when making this design decision?

Thanks.

kolusu · Posted: Thu Aug 12, 2004 7:34 am Post subject:

Fin,

Are you planning to load the entire vsam record into internal table? If so what is the LRECL of the vsam file? If you are using a serial search (SEARCH) verb then it is going to affect the performance. If the current design is using a vsam file, what is the access mode and the file organization(ksds,rrds,esds)

If you are going to use an internal table, how are you planning to load the internal table? Are you going to hardcode the values , or read it from a file?

Kolusu
_________________
Kolusu
www.linkedin.com/in/kolusu

Mike Chantrey · Posted: Thu Aug 12, 2004 10:30 am Post subject:

If it's a VSAM file you can load it up and it will be in key order, and you will be able to use the provided COBOL binary search facility (SEARCH ALL) which is very efficient - it should be <13 compares for locating any item by key, since 6000<(2^13) .

Fin · Beginner Joined: 19 Jan 2004 Posts: 27 Topics: 8

The VSAM file is created in one program and used (accessed) in another program
that immediately follows it.

The Logical Record length is 7 (consisting of 2 fields, 6-byte + 1-byte).
The space attribute is defined as: SPACE=(7,(600,50),RLSE)
Organization - indexed
Access mode - random

The file is sorted on the 6-byte field.
It is then read into the program and a compare is done on the 6-byte field against a
field thats being read in separately.

The proposed change would be to write a flat file instead of a VSAM file,
create a COBOL table in the second program and read the data from the flat file into this.
Does this seem like a sensible approach?
The number of records varies every cycle (usually in the region of 6000 - 7000) - I would be
declaring a table to occur the maximum (probably 8000) records.

Thanks in advance for any advice.

kolusu · Posted: Fri Aug 13, 2004 7:37 am Post subject:

Fin,

The proposed change seems fine, as the data is less and the file is also sorted. you can use a binary search (SEARCH ALL) since the data is sorted.As Mike mentioned earlier,it is very efficient as it takes only <13 compares for locating any item by key, since 8000<(2^13= 8192) .

Hope this helps...

Cheers

Kolusu
_________________
Kolusu
www.linkedin.com/in/kolusu

Fin · Beginner Joined: 19 Jan 2004 Posts: 27 Topics: 8

Kolusu, Mike
Many thanks!

Fin · Beginner Joined: 19 Jan 2004 Posts: 27 Topics: 8

Hi - just wondering if anyone can figure this out - after changing from VSAM to internal table the job took 3 times the CPU time to run...no other changes made, just change from VSAM to table.
Thanks.

powerhawk · Posted: Tue Nov 16, 2004 6:53 am Post subject:

Is the table sorted in ascending key field order? It must be! And you must use SEARCH ALL. The table is quite easy to sort also after it has been loaded from the VSAM file. I can give you code for that, not the best made but it works. I also recommend you to use a variable table using OCCURS DEPENDING ON, the SEARCH ALL instruction works better with varialbe tables, and use a huge max occurs value for the table like 20.000. And of course a nice error handling if the entire table is filled up during execution. I have code for all this if you want.

Fin · Beginner Joined: 19 Jan 2004 Posts: 27 Topics: 8

i'd really appreciate if you could show me the code...many thanks for the helpful advice.

powerhawk · Posted: Tue Nov 16, 2004 7:54 am Post subject:

Here is sample code:

powerhawk · Posted: Tue Nov 16, 2004 8:05 am Post subject:

All used variables are not defined in the sample above. I think you will understand how the code works anyway.

Fin · Beginner Joined: 19 Jan 2004 Posts: 27 Topics: 8

yes, i am able to follow and will build my code around your sample...thanks a million.

Cogito-Ergo-Sum · Posted: Wed Nov 17, 2004 11:40 am Post subject:

Just two things, if you think paranoid.

VSAM KSDS automatically ensures sorted records and non-duplicates. If you are 100% sure of non-duplicates and sorted sequence from the program wrting to the (proposed) flat file, then it is OK. Else, a SORT step may help.
_________________
ALL opinions are welcome.

Debugging tip:
When you have eliminated all which is impossible, then whatever remains, however improbable, must be the truth.
-- Sherlock Holmes.