When a large number of records are inserted into a relation in a short period of time, it is often recommended that all indices be dropped, and recreated after the inserts have been completed.

a. What is the motivation for this recommendation?

b. Dropping and recreation of indices can be avoided by bulk-updating of the indices. Suggest how this could be done efficiently for B+-tree indices.

c. If the indices were write-optimized indices such as LSM trees, would this advice be meaningful?