just my 2 ct

- with mass deletes:

- limit the number or deletes per transaction to avoid (dead) locking.

- perform full table(s) maintenance after the operation to optimize

space consumption and statistics.

- with mass updates:

- determine your impact scope. i. e. can you perform the update in

chunks (like with delete) having your data still symanticaly correct ?

- keep in mind, the log-overhead, io-overhead due to page splits

if the row nolonger fits in the original page, ...

- perform full table(s) maintenance after the operation to optimize

space consumption and statistics.

- maybe you're beter off creating a staging table, truncating

the original and insterting data according to the clustering key

sequence.

- prepare the operation, determine impact timeframe(s), involve your data consumers !