>> does anyone think dictionary managed tablespaces are ever a good >> choice any more? If you do, please let me know when and why << Yes, in test where we have limited space because you cannot limit the table/index size via maxextents under a locally managed tablespace. I still think that a DBA should be able to assign a maximum size to an object because for most of our objects there is a size where it the objects extends beyound this value something is wrong. That and the fact I have seen an infinite insert loop make it to production where the key was sequence generated. All you need is an unlimited sized table assigned to a tablespace with extendable files to fill the disk and ruin you weekend that you now have to spend trying to release storage. I have seen several posts where sites failed to use maximum sizes on their data files that would have usually prevented this issue, but being able to limit object sizes also has benefits. -----Original Message----- From: oracle-l-bounce@xxxxxxxxxxxxx [mailto:oracle-l-bounce@xxxxxxxxxxxxx] On Behalf Of Mark W. Farnham Sent: Thursday, December 17, 2009 7:21 PM To: fmhabash@xxxxxxxxx; haroon_a_qureshi@xxxxxxxxx; Oracle-L@xxxxxxxxxxxxx Subject: RE: ASSM and high volume concurrent inserts By DMT do you mean dictionary managed? That is certainly not required to use "manual" aka freelists segment space management. A locally managed tablespace can be ASSM or freelists in all versions I'm aware of. extent management dictionary extent management local autoallocate extent management local uniform [SIZE <size_clause>] are the extent management possibilities and segment space management auto segment space management manual are the segment space management possibilities. If you specify extent management dictionary, you get segment space management manual. (And does anyone think dictionary managed tablespaces are ever a good choice any more? IF you do, please let me know when and why - and avoiding a full export import to get rid of dictionary management of an existing SYSTEM tablespace I don't really count as a "choice.") You're probably confused by the statement "The segment_management_clause is relevant only for permanent, locally managed tablespaces" in the manual. That is true, but only because ASSM is not available for dictionary managed tablespaces. Regards, mwf -----Original Message----- From: oracle-l-bounce@xxxxxxxxxxxxx [mailto:oracle-l-bounce@xxxxxxxxxxxxx] On Behalf Of FMHabash Sent: Thursday, December 17, 2009 6:15 PM To: haroon_a_qureshi@xxxxxxxxx; Oracle-L@xxxxxxxxxxxxx Subject: RE: ASSM and high volume concurrent inserts Not sure what you mean by 'manually managed'. To manage FLists yourself, you need a DMT TS, which I believe can't do in 10g (for read/write). We had a similar problem on a high frequency LOB insert, which we resolved by eliminating a composite unique constraint. About 90% of BBW's that I've seen resulted from poor application design or run-time anomalies. Focus on application design and run-time behavior before you start thinking tablespace and segment attributes. -----Original Message----- From: Haroon A. Qureshi <haroon_a_qureshi@xxxxxxxxx> Sent: Thursday, December 17, 2009 12:46 PM To: Oracle-L@xxxxxxxxxxxxx Subject: ASSM and high volume concurrent inserts Hello, A client is running an application that does high volume concurrent inserts (> 200million rows). Performance degrades with high buffer busy waits. The tablespace is ASSM managed, so we can't change the freelists to tune it. I want to try moving to a manually managed tablespace and tune the freelists. But not sure if the client will go for that change, given the effort and timelines. Has anyone come across performance issues with ASSM and high volume inserts? Anyway of tuning it with ASSM? Thanks in advance, Haroon -- //www.freelists.org/webpage/oracle-l -- //www.freelists.org/webpage/oracle-l -- //www.freelists.org/webpage/oracle-l -- //www.freelists.org/webpage/oracle-l