MVSFORUMS.com

Sarangadhar

Hi all,
Some of my people are not recommending rexx to use in batch enviroment. Is there any solid reason for that? Can you tell the bottlenecks of using REXX in production env? Coz we are planning to use rexx to some extent wherever it is difficult to do with jcl/cobol.

Also if you find any whitepapers on bottlenecks of rexx, is very much useful.
_________________
Thanks

Sarangadhar · Posted: Mon Sep 11, 2006 9:21 pm Post subject:

Sorry, I was not specific I guess.

I am looking for bottlenecks of using REXX in production batch envronment. we are not using for any online functionality.

All opinions are welcome.
_________________
Thanks

semigeezer · Posted: Tue Sep 12, 2006 12:07 am Post subject:

What types of tasks are you planning to do and how much are you planning to do? If you are simply automating some frequent tasks or running Rexx CGI scripts on your web server, there should be no cause for concern. If you are processing millions of data records doing type conversions, or processing huge ISPF tables then the interpretive nature of Rexx could be a problem. There are probably very few performance guidelines for Rexx. The only 2 things you have to remember are that 1) Rexx is interpreted, not compiled, and 2) all data is stored internally as strings so both arithmetic and string operations can do lots of data conversion in normal processing. Also, as part of (2), you can run into performance problems because of how Rexx does comparisons with inexact operators like = ( it trims spaces before compares for example) but Rexx has operators like == that do not do this and theoretically should be faster. Bottle necks are really application specific. EXECIO can be memory intensive if used incorrectly (reading a whole file into a stem for example), using stems as associative arrays is powerful but a little slow, and calling external functions is probably a bad performer since the external function might need to be retokenized each time (maybe not). I'd say just write up a sample of what you want to do and test it. Don't expect COBOL or assembler performance, but ease of coding and maintenance can make up for a little more CPU or elapsed time.

Sarangadhar · Posted: Tue Sep 12, 2006 2:07 pm Post subject:

We will use for :
dynamic allocation/open/creation/read/delete of files from rexx (since name of the files are known at run time)
calling cobol programs from rexx
calling cobol-db2 programs from rexx
submit a jcl from rexx

And as we go on, may have to use for lots.. So need the bottlenecks of using rexx in mvs batch production env. So that we can try to avoid whenever there are roadblocks for rexx on performance wise and ease of debugging and finding the solution when abends or any problems occur.

Also one more doubt:
REXX is interpreted language, since it is not compiled one, we can't use for DB2 directly and large data processing and data co0nversions. But I saw some where telling the compiling the rexx. can i get clarified.
_________________
Thanks

ofer71 · Posted: Wed Sep 13, 2006 12:37 am Post subject:

You can purchase a REXX compiler from IBM.

O.
________
volcano vaporizers

Sarangadhar · Posted: Thu Sep 14, 2006 3:56 pm Post subject:

Am not talking about where to get the compiler.. Anyway I came to know that even compiled rexx code is not as efficient as COBOL in data processing. So we decided to use rexx only for allocations and tso/ispf calls; whereever we can't automate things with COBOL/JCL.
_________________
Thanks