We will be merging three of our View databases into one.
I have questions concerning allocations for index and data extents.
Our current three databases have 8906 for the Index block size. The Data block size for the three databases is different for each – 3476, 3768 and 13682.
Cylinder allocations per extent range anywhere from 400 to 3000 cylinders for the Index and anywhere from 2400 to 8000 cylinders for the Data extents.
I created a database for testing the merge with space allocations of 4300 for both Index and Data and with block sizes of:
. Index – 8906
. Data – 32,760
The merged DB seems to be using more space than the three separate databases did combined.
Is there a formula available for calculating, or a recommendation, for the most efficient allocations, regarding cylinders per extent and block size?
Release : 14.0
Component : CA View
It was suggested that the client use a View database data extent blocksize of 13682, for maximum efficiency.
Note: The View database index extent blocksize of 8906 is the only suggested one to use.
The client found that using a data extent blocksize of 32760, in a test, used 89% of available space; whereas using a data extent blocksize of 13682 used only 49% of space available.
Regarding database extent sizing:
. View database extents that are 4,369 cylinders or less will be allocated as an IBM direct access data set.
Extents that are more than 4,369 cylinders will be allocated as physical-sequential data sets due to IBM restrictions.
. The View Systems Extension Data Set Security (SARXTD) feature can be used to apply security to the high-level qualifier of datasets, which can include both database extents as well as other View datasets (like CVDEPENU and CVDEMAC), depending on how you name them.
. There are a maximum of 255 database extents (255 data, 255 index) in a View database.
. As View database extents need to be allocated with contiguous cylinders, that could be a determining factor as to how the extents are allocated.
. There really is no product-determined way of saying as to which allocation type is "better", it would depend on what works best for your environment.