Anyone have on opinion on the usage of SECUREFILE LOBs for spatial data?

Question

Hi folks,
From everything I read the usage of SECUREFILE LOBs over BASICFILE LOBs appears to be a nobrainer.
http://www.oracle.com/technetwork/database/options/compression/overview/securefiles-131281.pdf
Yet the default table creation remains set to create BASICFILE LOBs so my 11g spatial data is still sitting in BASICFILE LOBs (also I need to push it to and from 10g for the moment).
Should we all be changing over to SECUREFILE LOB storage as a matter of course on 11g? Is there a big payoff for VARRAY storage beyond compression? A little payoff?
Oracle says SECUREFILE LOBs "dramatically improve performance". Hey, I want some of that!
http://www.oracle.com/newsletters/sap/products/database/oradb11g-features.html
But yet there just has not been much traffic on the topic.
Pro Oracle Spatial mentions weakly on page 250 that
"Secure File LOBs are expected to be faster than BASIC LOBs".
but seems to only be referring to the spatial index.
Here Godfrind says to use them with points clouds but only trumpets the compression aspect.
http://www.ncg.knaw.nl/Studiedagen/09PointClouds/presentations/PointCloud_14_AlbertGodfrind.pdf
For those of us using ArcSDE, its pretty easy to add the SECUREFILE LOB keyword to the dbtune keyword that governs creation of the spatial column. But not a peep on the topic from the ESRI folks that I can find.
Anyone have any experiences or opinions to share? I am hoping to finally leave 10g behind one day this year and putting together a migration plan. Should converting the LOBs be a first day bullet item or just something to consider sometime down the road?
Thanks!
Paul

Answer 1

I have been writing on the benefits of rounding ordinates on storage and performance.
That series has not yet been finished and fully published but my main finding (backed by stats/graphs) is that rounding ordinates does have an effect on performance as average coordinates per feature increases.
As part of the work I also looked at SECUREFILE LOB storage in the hope that some sort of zlib/rle etc compression of the sdo_ordinates would substantially reduce storage size and thus increase performance.
The reason I looked at this is because a customer has migrated from SDEBINARY to SDO_GEOMETRY (though most others are going from SDEBINARY to SDE.ST_GEOMETRY) and saw a substantial increase in storage costs. They also have taken a performance hit in some aspects of their operations but that may be due to ArcSDE programmer SDO_GEOMETRY access issues rather than any real issues with Oracle performance (though the perception is very much that Oracle is to blame). SDEBINARY was always a compressed storage format and with SDE.ST_GEOMETRY being WKB (of some description) also has a low storage footprint.
The result of my work on LOB storage, though not published, is that I agree with Paul: there is little benefit for a lot of pain.
However, I do still the whole question of storage size as being an area of benefit if reductions could be made as this is a necessary precursor for spatial being able to reap benefits if column-oriented storage ever migrates down to standard Oracle database.
regards
Simon