The DBC.Crashdumps database is an internal Teradata database initially installed by the DIPCRASH scripts when the system is installed. Because it is a Teradata database, it requires permanent space. Space is allocated and managed similar to other databases and can be modified using Teradata DML.
The size of the DBC.Crashdumps database is configurable and depends on available space and your administrative preferences. Just as the PDE dump directory should be large enough to hold approximately 4 to 5 raw dumps, the DBC.Crashdumps database should be large enough to hold 4-5 full crashdumps (a full crashdump is the total of all raw PDE dumps from each node).
By default the database is NO FALLBACK, and no dumps are saved if an AMP goes down. If you specify FALLBACK, you gain fault tolerance, but the crashdump will take significantly more time and resources.
Crashdumps space is allocated from current PERM space for DBC. Reducing crashdumps space makes more free space available in DBC. Increasing Crashdumps space reduces DBC space.