|
View previous topic :: View next topic
|
| Author |
Message |
Ken Nieman
New User
Joined: 22 Feb 2008
Posts: 2
Location: Chicago, Illinois
|
|
|
|
Currently using HIDAM (OSAM with 3500 byte Root-Only) database. On a nightly basis the HIDAM database is /DBR’ed and split into smaller databases using home grown COBOL routines. Each of the smaller databases are processed sequentially then merged into a single database prior to bringing up the online. Each of the smaller databases contain appox. 200k records and are processed DLI with DBRC and IRLM turned off. Each of these jobs completes using around 30 CPU seconds.
After converting the same HIDAM database to HALDB, the CPU times have increased dramatically, from 30 CPU seconds to 140 CPU seconds. This is an increase of over 450%. We expected to see an increase in CPU Time since we converted the jobs to execute BMP vs. DLI but not of this magnitude. We recently modified the OSAM Buffer Pool for the IMS Control Region/DLI SAS address space to match OSAM Buffer Pool from the DLI job. The overall performance of the BMP jobs showed little improvement.
Has anyone experienced similar issues or have any other suggestions? Thanks in advance for your assistance.
Ken |
|
| Back to top |
|
 |
Bitneuker
CICS Moderator
Joined: 07 Nov 2005
Posts: 1104
Location: The Netherlands at Hole 19
|
|
|
|
| BMP's consume more. Did you have a look the checkpoints? |
|
| Back to top |
|
|
Sandy Zimmer
Active Member
Joined: 13 Jun 2007
Posts: 826
Location: Wilmington, DE
|
|
|
|
| May I ask why you converted? |
|
| Back to top |
|
|
Bitneuker
CICS Moderator
Joined: 07 Nov 2005
Posts: 1104
Location: The Netherlands at Hole 19
|
|
|
|
| Sandy Zimmer wrote: |
| May I ask why you converted? |
Since a number of years shops demand there programs to run as BMP including checkpoint/restart. But let's wait for the answer. |
|
| Back to top |
|
|
Sandy Zimmer
Active Member
Joined: 13 Jun 2007
Posts: 826
Location: Wilmington, DE
|
|
|
|
| George - I understand that. BUT.....they are obviously creating files as they are purging info, so BMP would not be necessary? |
|
| Back to top |
|
|
Ken Nieman
New User
Joined: 22 Feb 2008
Posts: 2
Location: Chicago, Illinois
|
|
|
|
| Checkpoints are issued every 2000 updates. The jobs that split and merged the smaller databases into a single HIDAM are the most expensive jobs we have in the shop. These jobs also were also starting to impact our ability to meet earlier online availability SLA's. |
|
| Back to top |
|
|
Bitneuker
CICS Moderator
Joined: 07 Nov 2005
Posts: 1104
Location: The Netherlands at Hole 19
|
|
|
|
| There's a lot of info regarding HALDB over here. Search for HALDB and explore, you might find the solution. |
|
| Back to top |
|
|
Bitneuker
CICS Moderator
Joined: 07 Nov 2005
Posts: 1104
Location: The Netherlands at Hole 19
|
|
|
|
| Ken, you've got PM. |
|
| Back to top |
|
|
|
|
