If I may suggest ,

1. If you provide a test case ( with the create …. exec plans ) , that will be great .
2. This is not a place for personal attacks ( like .. mean sprited ,,,, ) .
3. In a scientic field , we would learn to agree on disagreeing .

Miladin

>> There are very few software products that really are 100% tested.

Yeah, that’s true, and it;s not Oracle’s fault. It would take decades to test each and every nuance.

However, blocksizes are fundamental . . .

>> That is a bold statement to make with out any references.

No problem, here are references that recommend multiple blocksizes:

http://www.dba-oracle.com/t_multiple_blocksizes_summary.htm

Metalink Note:46757.1 titled “Notes on Choosing an Optimal DB BLOCK SIZE” says that there are some benefits from having larger blocksizes, but only under specific criteria (paraphrased from Metalink):

- Large blocks gives more data transfer per I/O call.

- Larger blocksizes provides less fragmentation (row chaining and row migration) of large objects (LOB, BLOB, CLOB)

- Indexes like big blocks because index height can be lower and more space exists within the index branch nodes.

- Moving indexes to a larger blocksize saves disk space. Oracle says “you will conserve about 4% of data storage (4GB on every 100GB) for every large index in your database by moving from a 2KB database block size to an 8KB database block size.”

>> the [Oracle] software … was not properly tested

I guess that depends on one’s definition of “properly”. There are very few software products that really are 100% tested. Given the number of features and options that are available in a product like Oracle database, it is quite likely they all do not get tested. Obviously when bugs are worked around, (as in this case) and not diagnosed, they do not get resolved as quickly either.

>> [re only 8k block testing]

It is doubtful that only an 8k block is tested, but since 8k is the default I think it probably gets tested more than a non-default value. I would assume that to be the case with every option though: the default value is tested more frequently than non-defaults.

>> given that the Oracle documentation, Oracle University and MetaLink all recommend non-standard blocksizes under special circumstances

That is a bold statement to make with out any references. I believe you may be reading too much into the Oracle documentation, etc. I have yet to see a recommendation on a non-default block size. I believe the documentation mentions there may be certain scenarios that a non-default block size may alter performance, but I do not believe I have ever seen a specific recommendation. Do you have links to the official documentation you are paraphrasing to support this observation? What is the exact verbiage? What are Donald K. Burleson’s recommendation on block sizes? I’ve seen one page of a few with some points, but these comments do not state much other that the simple points or comments from other people. Just for one example, one remark you have is “Large blocks gives more data transfer per I/O call”. True statement, but what about when multiblock read is used to read more than one block. Then what? If I may, this is like saying a larger petro tank holds more fuel than a smaller one. Completely true, but what does that do for fuel efficiency? I hope that you can elaborate on your position on block sizes. It’s interesting but I have not seen much details to persuade me to consider a non 8k block size. Also, the points that you have on your block size page (faster updates, etc): are those your findings or are they a collection from others? Thank you much for helping to understand more.