View previous topic :: View next topic
|
Author |
Message |
dejunzhu
Active User
Joined: 08 May 2008 Posts: 390 Location: China
|
|
|
|
Because batch programs and online programs are updating the same tables, I want batch program commit changes more when less online programs are updating , and batch program commit less when more online programs are updating. Thus, lock can be heavily avoided .
I wonder if there is a technique called such 'dynamic commit'?
If yes, how? |
|
Back to top |
|
|
dbzTHEdinosauer
Global Moderator
Joined: 20 Oct 2006 Posts: 6966 Location: porcelain throne
|
|
|
|
no, there is no such thing as dynamic commit.
there is such a thing as decent design.
what is your lock level? page?,row?
what is your commit frequency for batch?
how many updates are you making in cics before you syncpoint
or terminate the transaction?
you mentioned before that you are/have 'batchitzed' your cics system.
have you given any thought to having the 'online-batch processes'
redesigned so as to process them in batch environment? |
|
Back to top |
|
|
enrico-sorichetti
Superior Member
Joined: 14 Mar 2007 Posts: 10873 Location: italy
|
|
|
|
The commit frequency is completely free to choose.
the only thing to take care is to have enough info for a restart
( basically to reposition correctly input and output - non DB2 - entities )
but why change it dynamically, choose a reasonable one to start with and stick to it |
|
Back to top |
|
|
dbzTHEdinosauer
Global Moderator
Joined: 20 Oct 2006 Posts: 6966 Location: porcelain throne
|
|
|
|
problem with the word dynamic is that it implies/means
that some program/system is making decisions on a real-time basis.
you could have for each job/transaction
a db2 row that indicates level 1,2,3,4 (3 is usually enough) commit frequencies.
another table that contains level indicator for job/transaction.
at each commit or syncpoint, the job/transaction refreshes its copy of the commit level indicator,
and adjusts the commit/syncpoint frequency accordingly.
the commit level indicator would be maintained by dba's monitoring the system processing.
i have written syb-systems that follow this type of design,
and as long as you have good monitoring tools,
and the dba's are experienced with the work flow,
so that can anticipate - based on knowledge - when the system will probably incur stress points
and react accordingly,
it sorta works.
i imagine you could later write a subsystem to replace the dba's activities
with respect to monitoring and reacting - setting commit level indicator.
that is about as dynamic as you could get,
remembering that the last part can get you into trouble
if not written well,
/start personal opinion/
sort of like electing democrats to US congress and president.
/end personal opinion/ |
|
Back to top |
|
|
prino
Senior Member
Joined: 07 Feb 2009 Posts: 1306 Location: Vilnius, Lithuania
|
|
|
|
dbzTHEdinosauer wrote: |
/start personal opinion/
sort of like electing democrats to US congress and president.
/end personal opinion/ |
Which is still 100% better than letting the Koch brothers buy the presidency to feather their own interests. |
|
Back to top |
|
|
dbzTHEdinosauer
Global Moderator
Joined: 20 Oct 2006 Posts: 6966 Location: porcelain throne
|
|
|
|
prino,
who would you rather have buy the presidency to feather their own interests(rep) or pockets(dem)? |
|
Back to top |
|
|
dick scherrer
Moderator Emeritus
Joined: 23 Nov 2006 Posts: 19244 Location: Inside the Matrix
|
|
|
|
Hello,
Quote: |
Is there a technique called 'dynamic commit'? |
No, and there is no need for one. . .
If proper design was done and then proper implementation, there is No need for such a thing. If those designing/implemeting this do not understand how to build a system that easily supports both batch and online processing, someone who does should be brought in to advise.
No matter what someone clueless dictates (as far as design / implementastion) there are ways to get Major systems running that do both successfully.
[consternation]
The farther this development goes, the more it sounds like it will be problematic (if even supportable) when completely implemented with all intended fuctions/users.
[/consternation] |
|
Back to top |
|
|
dejunzhu
Active User
Joined: 08 May 2008 Posts: 390 Location: China
|
|
|
|
Thanks, experts,
I'll think over that. your suggestions are very valuable. |
|
Back to top |
|
|
|