Cost of Index Skip Scan

         0          0          0  DELETE  MTL_SYSTEM_ITEMS_INTERFACE (cr=676 pr=457 pw=0 time=761695 us)
       806        806        806   INDEX SKIP SCAN MTL_SYSTEM_ITEMS_INTERFACE_N4 (cr=640 pr=0 pw=0 time=6248 us cost=277 size=6930 card=33)(object id 1989058)

I decided to create a separate index just on the SEGMENT1 column, MTL_SYSTEM_ITEMS_INTERFACE_T, to see if that would improve the response time. The explain plan is shown below Rows (1st) Rows (avg) Rows (max) Row Source Operation
---------- ---------- ---------- ---------------------------------------------------

         0          0          0  DELETE  MTL_SYSTEM_ITEMS_INTERFACE (cr=3 pr=0 pw=0 time=37 us)
       806        806          0   INDEX RANGE SCAN MTL_SYSTEM_ITEMS_INTERFACE_T (cr=3 pr=0 pw=0 time=12 us cost=1 size=2743 card=13)(object id 7755594)

There is a marked improvement in the cost of index scan when doing regular index scan. We normally do not create custom indexes on standard Oracle tables but since this statement is part of a one-time conversion run that we are trying to optimize to fit in the allotted outage window, we can create this index temporarily and then drop it after the conversion run.

I am trying to understand why the cost associated with the SKIP SCAN would be that much different than a regular scan.

Thanks,
Amir

--
http://www.freelists.org/webpage/oracle-l