Access Methods:

Given two relations sized 400 pages each, and 20 pages of memory buffer,
compute the cost of a hash join.  Assume that attribute values are evenly
distributed, and that you can achieve 80% occupancy in each hash bucket on
average before any bucket overflows.  Assume a page size of 4KB 
and a bucket label and pointer structure size of 32B.

Use 20 pages of memory buffer to perform an external mergesort on a relation
of size 1000 pages.  Compute the I/O cost.

In the problem above, suppose that each tuple in the relation is 128 bytes,
but only 32 bytes are retained after a projection.  Compare the cost of
the external sort performed before and after the projection.

Consider a selection predicate as follows:
((Age < 25) AND (Sal > 30,000))  OR  ((Age > 25) AND (Sal > 100,000))
The selectivities of the predicates are 0.1, 0.6, 0.8, and 0.01 respectively.
How would you evaluate this selection if you have available to you
(a) Only a clustered index on Age.
(b) Only a clustered index on Sal.
(c) Separate indices on Age and on Sal, clustered on Age.
(d) Separate indices on Age and on Sal, clustered on Sal.

Exercises 14.2 through 14.5 from the cow book.

Work out all the examples given in Sec. 8.4 of the cow book -- but let C (the
CPU cost of processing one record) be zero.