I'm thinking about making the switch to LMT one of these days.
Can anybody think of valid (or not so valid) reasons for not doing this?

Another question I'd like to run by the congregation is if it would be beneficial to use multiple tablespaces in conjunction with LMT, so that we can group tables by different extent sizes (say 1MB, 10MB, 100MB)?

Is there even any discussion on this subject? I just don't see any disadvantages.

Gaja Vaidyanatha says of LMT's:
"LMTs eliminate file-level free space fragmentation issues. They also eliminate some object-level fragmentation issues related to extents of different sizes. LMTs manage the extents of an object by using either UNIFORM or AUTOALLOCATE options for tablespace space management. The space management for LMTs is done using a bitmap, in the first extent of the tablespace. Not curing the disease (PCTFREE and PCTUSED) will cause block-level and row-level fragmentation even for objects stored in LMTs. And in the bigger scheme of things, block-level and row-level fragmentation is what counts against performance."

As for grouping into different sized tablespaces, both Gaja and Oracle are advocating this approach as a best practice. Refer to the article from Oracle titled: "HOW TO STOP DEFRAGMENTING AND START LIVING: THE DEFINITIVE WORD ON FRAGMENTATION" by Bhaskar Himatsingka, and Juan Loaiza Oracle Corporation. Gaja's book Oracle 101 Peformance Tuning advocates a four bucket sizing of tablespaces as a means of reducing space fragmentation.


Gilbert Guymer
Database Administrator
Lufkin Industries, Inc.

I think that LMTs are usefull for more than 90% of the Baan Oracle users.

As Francescoalready mentioned fragmentation is a thing of low priority today. More coonsern about things like:
- spreading the load on different spindles
- backup/ recovery of tablespaces/ companies

Kind regards,
Dick