Performance

Attempts to benchmark veil using pgbench have not been very successful. It seems that the overhead of veil is small enough that it is overshadowed by the level of "noise" within pgbench.

Based on this inability to properly benchmark veil, the author is going to claim that "it performs well".

To put this into perspective, if your access functions do not require extra fetches to be performed in order to establish your access rights, you are unlikely to notice or be able to measure any performance hit from veil.

If anyone can provide good statistical evidence of a performance hit, the author would be most pleased to hear from you.

Next: Credits