What do almost fifty thousand Oracle objects look like?
Continuing my exploration of graph databases, I loaded every object from DBA_OBJECTS into a NEO4J graph and visualized it with Gephi.
That included all the objects from Oracle 11.2 DBA_OBJECTS, with the exception of the Java objects: Java Source, Java Class, Java Data. It also included some other schemas I’ve loaded into the database, such as Perfstat, SH, SCOTT, BI, etc. Altogether, 48,690 objects, and 61,710 relationships were inserted into NEO4J and then imported into Gephi.
Good Data Source:
As I did this, I thought that DBA_OBJECTS makes a rather good dataset to experiment with. It’s freely available to any DBA. The data is not sensitive. There is lots of data: tens of thousands of rows. The relationships between the objects are listed in DBA_DEPENDENCIES. Most all the data points represent a connected tree structure. This is exactly what NEO4J and Gephi work well with.
Gephi has a number of layouts to work with. I used the OpenOrd layout in Gephi to visualize all the data. OpenOrd completed the layout quickly, in a few minutes. By contast, the Fruchterman Reingold layout did not complete even after I let it run all night. Perhaps Fruchterman Reingold is only good for small data sets.
I partitioned according to the Object_Type (Synonym, View, Index, Table), using the default colors Gephi provided. Then set the edge color to red. See what the results looked like.
The Big Picture:
OpenOrd showed stelliums of objects clustered together.
DBA_OBJECTS Visualized In OpenOrd Layout, wide
They remind me of constellations in the night sky.
Read the rest of this entry »