How To: Estimate raster storage requirements for ArcSDE on Oracle


Intructions provided describe how to estimate raster storage requirements for ArcSDE on Oracle. Prior to importing large amounts of raster data into ArcSDE, it can be important to know in advance how much storage space is required. Storage requirements for ArcSDE rasters are influenced by several factors, including the color depth; number of bands; and loading parameters, such as tile size, compression, and reduced-resolution pyramids. Guidelines are provided for estimating raster storage requirements for ArcSDE on Oracle.


ArcSDE provides two methods for compressing image data to save space and improve performance: LZ77 and JPEG. The degree of compression depends on the distribution of pixel values in the image. Images with larger areas of similar colored pixels generally compress better than more heterogeneous images. To estimate storage requirements for ArcSDE rasters, it is necessary to load one or more sample images that are representative of the entire set in terms of size and heterogeneity and examine their storage usage in Oracle.

  1. Select a representative image and import it into ArcSDE using the same storage parameters that are used when importing the entire set of images.
  2. Analyze the raster table by right-clicking on the raster in ArcCatalog and selecting Analyze. Leave the checkbox for 'Raster Table' checked in the dialog box and click OK.
  3. Obtain the rastercolumn_id value from SDE.RASTER_COLUMNS table for the image that was just imported. Log in to SQLPlus as either the owner of the raster or as the SDE user and run the following query:
    SQL>SELECT rastercolumn_id from sde.raster_columns where table_name = ";"

  4. Run the following query using the value for the rastercolumn_id that was obtained in the step 3.
    SQL>SELECT SUM(bytes) / 1048576 size_in_megs FROM user_segments WHERE segment_name = 'SDE_BLK_<raster column id>';

  5. Multiply the value obtained in the previous step by the number of images to be imported.