IBM Mainframe Forum Index
 
Log In
 
IBM Mainframe Forum Index Mainframe: Search IBM Mainframe Forum: FAQ Register
 

Guidelines for analyzing Strobe result


IBM Mainframe Forums -> Compuware & Other Tools
Post new topic   Reply to topic
View previous topic :: View next topic  
Author Message
raghavmcs

Active User


Joined: 14 Jul 2005
Posts: 105

PostPosted: Fri Dec 04, 2009 8:49 pm
Reply with quote

Last night I strobed one of the most MIPS consuming batch job in production.This is one of the most critical batch update job running during nightly primary window uses mostly VSAM KSDS files and multiple called programs.We have macros to perform i/o operation on VSAM.After discussing with some help around me I strobed with 10000 sample size the cpu time on that day was around 54.23.
I was scanning the strobe results which is around 6k lines long and see a lot of stats is captured.Unfortunately I can not post that here in forum as its against guidelines.
I strobed this process to explore any improvements chances if any as its one of the top CPU consuming job which runs dailly.
I am wondering for what possible places I should explore?are there any guidelines to go ahead step by step.

At the same time I was thinking for when initially mainframe files/critical program are designed/defined to meet the initial needs and by the time when business expands(usually it expands if its well managed and economy condition is good) are there any possibilities of finding scope of improvement.Here I am assuming for if business has seen the long life cycle of around 30 years.
Please provide your expert advise from your experience.My experience is not very vast.
Back to top
View user's profile Send private message
Robert Sample

Global Moderator


Joined: 06 Jun 2008
Posts: 8696
Location: Dubuque, Iowa, USA

PostPosted: Fri Dec 04, 2009 8:56 pm
Reply with quote

Your best bet would be to work with coworkers experienced in reviewing STROBE output, or contacting Compuware for assistance in interpreting the output. If you can't give us anything to look at, all we could do is guess and most likely guess wrong at that.

It sounds like spending some time in the STROBE manuals would be a good use of time, however.
Back to top
View user's profile Send private message
raghavmcs

Active User


Joined: 14 Jul 2005
Posts: 105

PostPosted: Sat Dec 05, 2009 12:49 am
Reply with quote

Dear Robert,

Thanks I totally agree with you but I do not get much support from other as all are busy with projects.

While looking at the Strobe report I see one of the section which gives sectionwise information about hits on various VSAm file for read,update,delete operations.

If I post the information from the Vsam section will it be possible for if something can be suggested out of that information.I think posting information from that section is not going to put me in any issues...thanks
Back to top
View user's profile Send private message
Robert Sample

Global Moderator


Joined: 06 Jun 2008
Posts: 8696
Location: Dubuque, Iowa, USA

PostPosted: Sat Dec 05, 2009 12:58 am
Reply with quote

We can look it -- until we see it, no guarantees, but it'll be worth looking at.
Back to top
View user's profile Send private message
dick scherrer

Moderator Emeritus


Joined: 23 Nov 2006
Posts: 19244
Location: Inside the Matrix

PostPosted: Sat Dec 05, 2009 1:14 am
Reply with quote

Hello,

Quote:
but I do not get much support from other as all are busy with projects.
If this is "one of the most critical" nightly batch jobs, i suggest the management might make some of their time avaliable if they could help. . .

To say it is "most critical" yet no one that can help is to look at it because they are "too busy" does not sound right. . . icon_confused.gif
Back to top
View user's profile Send private message
Akatsukami

Global Moderator


Joined: 03 Oct 2009
Posts: 1788
Location: Bloomington, IL

PostPosted: Sat Dec 05, 2009 1:20 am
Reply with quote

dick scherrer wrote:
Hello,

Quote:
but I do not get much support from other as all are busy with projects.
To say it is "most critical" yet no one that can help is to look at it because they are "too busy" does not sound right. . . icon_confused.gif

You've spent how many years in production support without hearing this? icon_lol.gif
Back to top
View user's profile Send private message
raghavmcs

Active User


Joined: 14 Jul 2005
Posts: 105

PostPosted: Sat Dec 05, 2009 3:08 am
Reply with quote

Thanks I will try to post something out of it.First I am looking at the detailed reports.

Answer to Akatsukami,

We have 26 alpha which can make n possible words(good bad motivative .......) and many of them I never heard many might be stil under list of new addition..

I beleive learning is continual process...I never learnt mainframe in my basics and never aspired to :-) Cheers
Back to top
View user's profile Send private message
Marso

REXX Moderator


Joined: 13 Mar 2006
Posts: 1353
Location: Israel

PostPosted: Sat Dec 05, 2009 3:34 pm
Reply with quote

I've seen some STROBE reports, it's not always obvious to find something.
I usually look for peaks and big numbers...

VSAM access can be tuned by use of the AMP parameter.
So you can try some options and compare the results with STROBE. If you can pinpoint one of the "busiest" files and enlarge its buffers (for example), you may achieve a significant difference.

To be effective, you should know which kind of access you use for each file (random or sequential).
Then you can try these parameters accordng to these rules (see Acquiring Buffers).

Have a look at the LISTCAT of these VSAM files too. Some may need to be reorganized ?
Back to top
View user's profile Send private message
MBabu

Active User


Joined: 03 Aug 2008
Posts: 400
Location: Mumbai

PostPosted: Sun Dec 06, 2009 12:35 am
Reply with quote

In my experience (very limited) both STROBE and APA (IBM's equivalent) require an in depth knowledge of both the application program you are analyzing and MVS (z/OS) itself. The reason is that the performance problems come in two flavors: environment and algorithm. Environment problems are things like buffers, block sizes, work load manager settings, etc. But by the time you get to using STROBE or APA, you should be past those basic elements and down to the more crucial question... is the algorithm in your program as efficient as it can be? STROBE will tell you how much time is spent in I/O routines or waits or code blocks but that is not always, in itself, useful.

Here is an example of what I mean. I worked on a problem where a source code management system was taking 15 minutes to do something seemingly trivial. IBM's performance group had looked over this for a couple of weeks looking at I/O usage and the usual suspects and could not identify the problem. I know a little about application's internals so I got the APA report and saw that out of those 15 minutes, there was a call to a particular module that took a total of 0.034 seconds. Because I knew the application, I suspected that that module should not be called at all and indeed the problem was that it was being called repeatedly and it called other programs that did lots of unnecessary PDSE member listing but looked like regular I/O in the APA report. After a 2 line tweak to the application, it went from 15 minutes to a few seconds. The point is, unless you see huge blocks of time spent in your application code itself, you will need to understand the application. Knowing what z/OS routines do what function also helps because you can tell what types of I/O operations are being done.
Back to top
View user's profile Send private message
dick scherrer

Moderator Emeritus


Joined: 23 Nov 2006
Posts: 19244
Location: Inside the Matrix

PostPosted: Sun Dec 06, 2009 4:23 am
Reply with quote

Hello,

Echoing what MBabu says, the drawback with resource monitors is that there is not a way for the monitor to analyze that the process is properly designed for the task at hand.

The "worst case" i'll mention here happened many years ago when my client was implementing a replacement of their customer service/order entry system. To avoid repeating validation code in multiple modules, they went to "common subroutines" for editing everything and these modules were available to both batch and online processes. All of the online was sub-second and high-volume batch update tests ran quite acceptably. So far, so good. . .

The decision was made to use the same subroutiones for the data conversion programs to migrate the "old" data to the new database. Unfortunately, even a high-volume batch test run was nothing compared to full-file conversion (a few thousand per batch run and 40 million or so to convert). After the conversion team determined that the full-file conversion would take between 6 and 7 days to run, i was asked if anything could be done to improve the time used by the subroutines. With 2 of the conversion team to explain their process, i decided that the 40+ validation tables (most with less than 50 entries) should not be referenced in the conversion code (and got into a bit of hot water when i suggested that the subroutine use be discontinued).

We cloned the conversion code and used in-core arrays rather than database tables and after a bit of other re-structuring were able to run the conversion in under 3 hours.

Suggest someone take a long look at what the process is actually doing (rather than how it is doing it). If the approach is self-defeating, looking at low-level specifics won't help much.
Back to top
View user's profile Send private message
raghavmcs

Active User


Joined: 14 Jul 2005
Posts: 105

PostPosted: Tue Dec 08, 2009 12:10 am
Reply with quote

Thanks Experts,for all your valuable thought.I think I have to take this to the next level of solution team experts here.Once I am done with filtering some results from strobe I would post them here.
Back to top
View user's profile Send private message
dick scherrer

Moderator Emeritus


Joined: 23 Nov 2006
Posts: 19244
Location: Inside the Matrix

PostPosted: Tue Dec 08, 2009 12:14 am
Reply with quote

Good luck icon_smile.gif

We will be interested to see what is learned,

d
Back to top
View user's profile Send private message
MBabu

Active User


Joined: 03 Aug 2008
Posts: 400
Location: Mumbai

PostPosted: Tue Dec 08, 2009 9:55 am
Reply with quote

Any observations you gain would be helpful to others (me) so if you find any tips, we'd welcome them (hope I can speak for the group here)
Back to top
View user's profile Send private message
avik1983

New User


Joined: 11 Mar 2010
Posts: 5
Location: Chennai

PostPosted: Thu Mar 25, 2010 3:54 pm
Reply with quote

I think I am replying too late in this question,but hope it will help someone after.

In strobe many details r there.I will put a small doc here...Hope this is useful
Back to top
View user's profile Send private message
dick scherrer

Moderator Emeritus


Joined: 23 Nov 2006
Posts: 19244
Location: Inside the Matrix

PostPosted: Thu Mar 25, 2010 8:48 pm
Reply with quote

Hello and welcome to the forum,

Thank you for sharing your report icon_smile.gif

Is it possibly available via some web site? More than a few of our members are not permitted to download attachments.
Back to top
View user's profile Send private message
View previous topic :: :: View next topic  
Post new topic   Reply to topic View Bookmarks
All times are GMT + 6 Hours
Forum Index -> Compuware & Other Tools

 


Similar Topics
Topic Forum Replies
No new posts first column truncated in search result IBM Tools 13
No new posts SDSF like solution in EJES (store com... All Other Mainframe Topics 4
No new posts executing XCTL command in COBOL witho... CICS 10
No new posts Undesired result while joining files. SYNCSORT 3
No new posts COBOL - Array result in single variab... COBOL Programming 3
Search our Forums:

Back to Top