Portal | Manuals | References | Downloads | Info | Programs | JCLs | Master the Mainframes
IBM Mainframe Computers Forums Index
 
Register
 
IBM Mainframe Computers Forums Index Mainframe: Search IBM Mainframe Forum: FAQ Memberlist Usergroups Profile Log in to check your private messages Log in
 

 

Performance of SORT with and without blocksize

 
Post new topic   Reply to topic    IBMMAINFRAMES.com Support Forums -> JCL & VSAM
View previous topic :: :: View next topic  
Author Message
raghavmcs

Active User


Joined: 14 Jul 2005
Posts: 105

PostPosted: Fri Jan 30, 2009 3:55 am    Post subject: Performance of SORT with and without blocksize
Reply with quote

Dear Experts,

I was just doing the verification to get the conclusion in terms of performance for if we run with blksize or let the system decide the blksize.

the jcl I used for this is .I ran this with test.abc was having around 80k records

Code:

//STEP10   EXEC PGM=SORT,COND=(0,LT)                   
//SORTIN   DD  DSN=test.abc,DISP=SHR     
//SORTOUT  DD  DSN=test.abcd,       
//             DISP=(,CATLG,DELETE),                   
//             DATACLAS=LARGE,                         
//             DCB=(RECFM=VB,LRECL=2004,BLKSIZE=2008)   
//SYSOUT   DD  SYSOUT=H                                 
//SORTMSG  DD  SYSOUT=H                                 
//SYSIN    DD  DSN=test.PARMLIB(SRTXRP4),DISP=SHR       
----+----1----+----2----+----3----+----4----+----5----+-
  SORT FIELDS=(72,3,A,230,60,A,124,3,A),FORMAT=BI       
  OMIT COND=(23,5,CH,NE,C'XRP4 ')                       


For this the job iof is as follows

Code:

-------------------------------------------------------------------------------------
IEF373I    STEP/STEP10     /START 2009029.1703                                       
IEF374I STEP/STEP10  /STOP  2009029.1703 CPU    0MIN 00.19SEC SRB    0MIN 00.00S
IEF285I   SYS3.MVSV5.RUNTIME                           KEPT                     
IEF285I   VOL SER NOS= CATEEE.                                                 
IEF375I  JOB/CNJRP000/START 2009029.1703                                       
IEF376I  JOB/CNJRP000/STOP  2009029.1703 CPU    0MIN 00.19SEC SRB    0MIN 00.00S
-------------------------------------------------------------------------------------


After this I reran the jcl with removing the blksize and with the same volume of data.This time the job iof is as follows

Code:

------------------------------------------------------------------------------------
IEF373I    STEP/STEP10     /START 2009029.1702                                       
IEF374I STEP/STEP10  /STOP  2009029.1702 CPU    0MIN 00.20SEC SRB    0MIN 00.00S
IEF285I   SYS3.MVSV5.RUNTIME                           KEPT                     
IEF285I   VOL SER NOS= CATEEE.                                                 
IEF375I  JOB/CNJRP000/START 2009029.1702                                       
IEF376I  JOB/CNJRP000/STOP  2009029.1702 CPU    0MIN 00.20SEC SRB    0MIN 00.00S
-----------------------------------------------------------------------------------


I had one observation here which is contradictory for me
If we are going to leave upto the system to decide the blocksize it should take less time rather that if we code the blksize.The time here are .19(with blksize) and .20(without blksize)
May be possible I am wrong in my understanding.
Please clearify my doubt for following
a)What could be a better option coding bw coding the blksize option than to leaving up the system.

Please let me know incase I need to furnish any more information,thanks
Back to top
View user's profile Send private message

gcicchet

Senior Member


Joined: 28 Jul 2006
Posts: 1703
Location: Australia

PostPosted: Fri Jan 30, 2009 4:40 am    Post subject:
Reply with quote

Hi,

efficient blocksize will definitely improve elapsed time and save disk space, the time it takes to run is affected by many other things,
see
http://ibmmainframes.com/viewtopic.php?t=36431&highlight=elapsed


Gerry
Back to top
View user's profile Send private message
Frank Yaeger

DFSORT Moderator


Joined: 15 Feb 2005
Posts: 7130
Location: San Jose, CA

PostPosted: Fri Jan 30, 2009 5:01 am    Post subject:
Reply with quote

raghavmcs,

If you think its valid to draw conclusions about performance based on one job about which you haven't given any information (e.g. number of records, bytes sorted, etc), then I can easily counter your argument based on a job I ran with RECFM=VB, LRECL=2004 and BLKSIZE=2008 vs no BLKSIZE (-> SDB BLKSIZE=27998) and the same control statements you used.

Code:


                          EXCPs     CPU      Elapsed
With BLKSIZE=2008         1731    40.86          3.0
Without BLKSIZE (SDB)     1595    19.47          1.3



The lesson from this is - valid performance testing is an art.
Back to top
View user's profile Send private message
Robert Sample

Global Moderator


Joined: 06 Jun 2008
Posts: 7993
Location: Bellevue, IA

PostPosted: Fri Jan 30, 2009 5:24 am    Post subject:
Reply with quote

The original statement was 80K records ... but one test for performance measurement is pretty much useless. 10 to 20 tests I could see maybe making some inferences from, but using one job output to determine performance isn't much better than closing your eyes and throwing darts to pick which way to go.

Not only that but the original difference was .20 vs .19 seconds -- which could very easily be not a difference if fully investigated (.194999 vs .195001 seconds for example). Not to mention the impact site standards could have on the job -- overriding the block size, for example.
Back to top
View user's profile Send private message
Frank Yaeger

DFSORT Moderator


Joined: 15 Feb 2005
Posts: 7130
Location: San Jose, CA

PostPosted: Fri Jan 30, 2009 5:34 am    Post subject:
Reply with quote

Quote:
The original statement was 80K records


The statement that "test.abc had 80K records" was a bit meaningless - we don't know how many records were deleted by the OMIT, sorted or written to the output data set.

Of course, the rest of what you said is true.
Back to top
View user's profile Send private message
Robert Sample

Global Moderator


Joined: 06 Jun 2008
Posts: 7993
Location: Bellevue, IA

PostPosted: Fri Jan 30, 2009 6:39 am    Post subject:
Reply with quote

You're right, Frank, I forgot about the OMIT -- so the input 80K record count means nothing.
Back to top
View user's profile Send private message
View previous topic :: :: View next topic  
Post new topic   Reply to topic    IBMMAINFRAMES.com Support Forums -> JCL & VSAM All times are GMT + 6 Hours
Page 1 of 1

 

Search our Forum:

Similar Topics
Topic Author Forum Replies Posted
No new posts Innovations in Your Mainframe Sort Arun Raj PC Guides & IT News 4 Wed Mar 15, 2017 10:38 pm
No new posts SORT VSAM file with each field one by... maxsubrat DFSORT/ICETOOL 6 Tue Mar 14, 2017 1:07 pm
No new posts SORT requirement sjiraga COBOL Programming 6 Thu Mar 09, 2017 9:11 pm
No new posts SORT to append Month and YEAR in MMYY... tisamf DFSORT/ICETOOL 1 Wed Mar 08, 2017 4:46 pm
No new posts JCL sort Match and NoMatch rakeshsekar1987 SYNCSORT 2 Thu Feb 02, 2017 10:26 pm


Facebook
Back to Top
 
Mainframe Wiki | Forum Rules | Bookmarks | Subscriptions | FAQ | Tutorials | Contact Us