locally managed tablespaces performance

Hi ,

My system is a datawarehouse. I use locally managed tablespaces with extent size of 1MB on OCFS . I have large tables , and since these tables reside in locally managed tablespaces , with 1MB extent size , i see lots of extents allocated. For instance one of the tables has about 6000 extents .

 Is there any performance impact of having too many extents? I usually use parallel query and parallel query baypasses the buffer cache. But , when i monitor the buffer pool , i see some of the blocks are cached. It seems that these blocks are the first or second blocks of each extent allocated to the table . (And these are read by using db seq.scan not direct path read ) Why are these blocks read by using db seq scan and cached? I guess these are the bitmap blocks which defines the extent but i really wonder the truth. What happens internally when oracle reads a table? How does it clarify the blocks that are read ?

Does having a large number of extent impact performance ?

Kind Regards,
hope Received on Tue Oct 12 2004 - 02:03:08 CDT