Query Evaluation:

Give an example of ((A join B) join C) where the output of a sort-merge join
for (A join B) can be fed as input to a simple merge join with C, with no
intermediate sorting required.  A,B,C are relations -- you pick the
attributes and the join conditions to make the above example work out.

Given binary relations A with 40,000 pages, B with 400 pages, and C with 20
pages, and given a buffer pool size of 40 pages, propose a query plan to
evaluate A join B join C, where each equijoin involves the second column of
the left relation and the first column of the right relation.  Let the
estimated size of A join B, and B join C, both, be 2000 pages.
Determine the IO cost of your proposed query plan.
Write a paragraph qualitatively justifying your design choice.

In the problem above, suppose there is a selection condition applied to A
join B, reducing its size to just 10 pages, would that change your answer
above?  Justify.

You have determined, for a selection query Q, that you prefer a full relation
scan to an indexed access.  
Given a query Q' with selectivity greater than Q, what can you say about the 
relative cost of scan versus index access.
Given a query Q" with selectivity less than Q, what can you say about the 
relative cost of scan versus index access.


Suppose you have 100 pages of memory, and each page can hold 100 entries.
You wish to compute the join of two relations: R1, which has 300 pages, and
R2, which has 3000 pages.  Compute the I/O required for:
a. The Grace hash join algorithm
b. The hybrid hash join algorithm.