Raster Catalog Upgrade; Oracle 9i/ArcSDE 9.1 SP2 to Oracle 11g/ArcSDE 10

Question

Hi all,

I'm looking to upgrade a large raster catalog of Imagery data (600GB) to a new database on a new platform.
Current specs are;
Oracle 9i
ArcSDE 9.1 SP2

Migrating to;
New Itanium HP-UX 11.31
Oracle 11.2.0.1 (64 bit)
ArcSDE 10.0

The vector datasets we will be migrating using ArcCatalog into a new empty database across the network and a fresh reload will be done of our reference data. However, I am not sure how this should be done with the Imagery database; I know that incremental steps will be required as it is rather a chasmic jump but if anyone has any experience of doing this could you please help?

Other options of course are ImageServer and storing the data in FGBD but if the above can be done relatively straightforwardly it would be preferable to do that.

Thanks in advance
Katie

GovindarajanThillaividankan · Answer

Hi all,

I'm looking to upgrade a large raster catalog of Imagery data (600GB) to a new database on a new platform.
Current specs are;
Oracle 9i
ArcSDE 9.1 SP2

Migrating to;
New Itanium HP-UX 11.31
Oracle 11.2.0.1 (64 bit)
ArcSDE 10.0

Ms.Katie

I am about to start a similar  exercise. I have massive vector data of Europe in ArcSDE 9.1 on Oracle 10g. I want to migrate this to ArcSDE 10 on Oracle 11g. When i referred to ArcGIS help documentation, It says that I cannot connect from ArcMap or ArcCatalog 10 to ArcSDE 9.1 using a SDE connect or a Direct Connect as ArcGIS 10 does NOT support ArcSDE 9.1

I do not have ArcGIS 10 so I am not able to verify this.

Did u face any problems?

Secondly, Can Data Interoperability i.e. FME Extension for ArcGIS 10 connect to ArcSDE 9.1?

Pls guide me.

Regards

Govind

JakeSkinner · Answer

I would recommend creating a Mosaic Dataset rather than a raster catalog.  You will get much better performance than a raster catalog and this is core at 10.  Not only will it provide better performance, but it will also save you 600 GB of storage space in the geodatabase.  Here are some other helpful links on mosaic datasets:

http://www.esri.com/news/arcuser/0610/mosaicdataset.html

http://video.esri.com/tag/37/mosaic-dataset

VinceAngelo · Answer

Actually, you can jump that chasm using the 'sde2raster' and 'raster2sde' utilities of se_toolkit.

First, you'd have to create new empty catalog(s) using ArcGIS.  Then you'd need to export the
rasters (one tile at a time, using the '-r' option or a 'where' clause) with 'sde2raster'.  While it's
possible to "pipe" between 'sde2raster' and 'raster2sde', in my experience, it's better to batch
the process with discrete intermediate steps, so if you need to stop processing, you can start
up again without having to delete a row from the target database (there's also fewer games
to play when executing a 9.1 binary with 9.1 SDEHOME and a 10.0 binary with a 10.0 SDEHOME).
I'd highly recommend that you change the tile size for 8-bit rasters on import to a 96x96 tile size
(with '-t 96,96' ), or consider changing the compression type so that the nominal 8K page size
of Oracle 11g is honored (without chaining the BLOB in *every* row).  If you prototype the
process by exporting from the fourth tile level  ('-L 3'), it will take a fraction of the time for
"whole table" trials.

The one issue you'll have to watch for is a bug where partial tiles (those with NODATA near tile
boundaries) are exported so that the last non-NODATA value "bleeds" south on import.  I haven't
been able to reliably reproduce this issue since 9.3 came out (which is why it never got fixed)
and haven't run 9.1 in ages.  I have worked on some masking tools to "repair" broken images, 
but not an automation tool to look for bleeding values.

I did a number of these upgrades between 9.0/9.1 and 9.2/9.3 but it's been a long time since
it was last needed.  In the long run, the Image Server extension capability of ArcGIS and ArcGIS
Server caches have eliminated the need to implement rasters in databases, and disk sizes have
changed to the point that a 600Gb catalog rattles inside the average firewire drive, making the
database overhead to manage the storage wasteful.

Good luck.

- V