Hi there,
could someone help me with a question on locking granularity please? I am revising for an exam on database design and concurrency and am trying to answer this question: http://i43.tinypic.com/mjtnox.png
I know that the end answer will be that table level locking granularity is required instead of tuple level granularity.. but I just can't think of an example of where the tuple level granularity won't guarantee serialisability.
thanks for any help!

For example, tuple-level locks won't prevent another transaction from inserting a new Accounts tuple while the first transaction computes a total for that branch.