View previous topic :: View next topic |
Author |
Message |
munnabai Beginner
Joined: 17 Aug 2007 Posts: 3 Topics: 3
|
Posted: Tue Sep 04, 2007 10:26 am Post subject: Reg BUFNI and BUFND |
|
|
Hi,
Can any one gimme a simple example from which I can understand how the BUFNI and BUFND parameters are calculated for a VSAM datset.
I searched in the internet most of the examples given does not make me to understand the basics allocation correctly. Guide me guys...
Thanx,
Munna. |
|
Back to top |
|
|
jsharon1248 Intermediate
Joined: 08 Aug 2007 Posts: 291 Topics: 2 Location: Chicago
|
Posted: Tue Sep 04, 2007 11:35 am Post subject: |
|
|
Each BUFNx value corresponds to the number of buffers allocated for your VSAM dataset. The BUFNI are for index buffers, the BUFND are for the data buffers. Each buffer is large enough to hold 1 CI.
The default allocation for batch is 1 index buffer and 2 data buffers. For random processing, you'll want at least 1 index buffer for each index level and a handful of data buffers, probably 5 or less. For sequential processing, you'll want to load up on the data buffers, usually around 20 or 30. For skip sequential, use the 1 index buffer for each index level guideline and a high number for the data buffers, 20 or 30. You reach a point of diminishing returns, so you'll need to experiment with the buffer allocations to determine the most appropriate for each file.
There's a good presentation of this in the 'VSAM Demystified' book (2.6.13) when you follow the 'Quick Manuals' link at the top of the pages in this forum. |
|
Back to top |
|
|
munnabai Beginner
Joined: 17 Aug 2007 Posts: 3 Topics: 3
|
Posted: Wed Sep 05, 2007 4:05 am Post subject: |
|
|
Thank U.....sharon |
|
Back to top |
|
|
expat Intermediate
Joined: 01 Mar 2007 Posts: 475 Topics: 9 Location: Welsh Wales
|
Posted: Wed Sep 05, 2007 12:17 pm Post subject: |
|
|
Quote: | For skip sequential, use the 1 index buffer for each index level guideline and a high number for the data buffers, 20 or 30. |
Be very careful when allocating buffers to skip sequential processing because every time that you swap from randon to sequential and back to random, you will probably find that all of your buffers are refreshed, thus creating extra I/O within your program. _________________ If it's true that we are here to help others,
then what exactly are the others here for ? |
|
Back to top |
|
|
|
|