Name:_________________________________________________________________________
CMSC 661 – Principles of Database Systems Spring 2011, Handout 3
Sections covered since Exam II: 10.1 to 10.9, 11.1 to 11.6, 12.1 to 12.5.3, 12.7, 13.1 to 13.3.3,
14.1 to 14.10.
1. Identify each part (A through F) of the slotted-page structure used to store variable-length
records.
A B C D E F
A = # of entries and pointer to free space
B, C = pointers to records
D = free space
E, F = records
2. For the B+ tree below,
a. What is n?
5
b. How many values can be stored in non-leaf nodes?
┌
n/2┐ to n, which is 3 to 5
c. Would a new value fit into one of the existing leaf nodes?
Yes, there‟s room in both leaf nodes, which can hold ┌(n-1)/2┐ to n-1 nodes,
which is 2 to 4
d. In which block would the value "Frederick" be stored?
In block 2
e. In which block would the value "Arbutus" be stored?
In block 2
3. Consider the following schedule.
Time T1 T2
1 read(A)
2 write(A)
3 read(A)
4 read(B)
5 write(B)
6 write(A)
Is the schedule conflict serializable as <T1, T2>? Explain why or why not.
Yes. You can swap instructions at times 3 and 4 (they work on different objects).
You can then swap instructions at times 4 and 5 (they work on different objects).
This results in <T1, T2>.
Although it wasn‟t part of the question, note that it is NOT conflict serializable at
<T2, T1>. To do this, you would need to move the instruction at time 3 up to time 1.
However, you can‟t swap the instructions at time 2 and 3, because they affect the
same object and at least one of them is a “write”.
4. Consider the following SQL query using R (A, B, C, D) and S (D, E), where r(R) and s(S).
Consider the relational algebra expression, derived from the SQL. Using equivalence rules,
suggest an equivalent relational algebra expression that is more efficient.
select s.E, r.A
from r, s
where r.D = s.D
and s.D = ‘Baltimore’
and r.B = ‘Sales’
s.E, r.A ( r.D = s.D s.D = ‘Baltimore’ r.B = ‘Sales’ (r s))
There are a number of solutions. Consider our rules of thumb, and perform the
select BEFORE the cross product.
s.E, r.A ( r.D = s.D (r.B = „Sales‟ (r) s.D = „Baltimore‟ (s) ))
2
5. Given R (A, B, C, D) and S (D, E), where r(R) and s(S), assume that r has 1,000,000 rows
with 100 rows stored per block, s has 100,000 rows with 500 rows stored per block, the block
seek time is 0.4 microseconds, and the block transfer time is 0.1 microseconds, there is a
primary index on A and a secondary index on B. Assume the height of any index used is 5.
Assume there are 5 rows where B = ‘Baltimore’. How long will it take to execute the
following statements?
B = ‘Baltimore’ ( r)
You need algorithm A5 for secondary index, equality on non-key.
(index height + number of rows) * (seek time + transfer time)
(5 + 5) * (0.4 + 0.1) * 10-6 seconds
= 5 * 10-6 seconds
r s
The Index-Nested-Loop join won‟t work, since there‟s no index on the join attribute
(D). This leaves us with Nest-Loop or the Block-Nested-Loop joins.
Here‟s the answer for the Block-Nested-Loop join.
To retrieve data from r, we need -- br * tS + br * tT
To retrieve data from s, we need -- br * tS + br * bs * tT
Note that there are 10,000 blocks in r and 200 blocks in s.
10,000 * 0.4 * 10-6 + 10,000 * 0.1 * 10-6 + 10,000 * 0.4 * 10-6 + 10,000 * 200 * 0.1 * 10-6
= 0.209 seconds
Out of curiosity, is it faster than the Nested-Loop join? Let‟s see ...
To retrieve data from r, we need -- br * tS + br * tT
To retrieve data from s, we need -- nr * tS + nr * bs * tT
Note that there are 10,000 blocks in r and 200 blocks in s.
Note that the Nested-Loop join uses nr as a multiplier.
10,000 * 0.4 * 10-6 + 10,000 * 0.1 * 10-6 + 1,000,000 * 0.4 * 10-6 + 1,000,000 * 200 * 0.1 * 10-6
= 20.405 seconds
