NetID: CS411 Midterm Exam

CS411 Database Systems

Fall 2008

University of Illinois at Urbana-Champaign

Midterm Examination
October 14, 2008
Time Limit: 75 minutes
• Print your name and NetID below. In addition, print your NetID in the upper left corner of every
page.

Name: NetID:

• Closed notes; closed book; no sheet of formulas permitted.

• Please write your answers directly on the exam sheet. The space we left for your answers is often more
than what you actually need. Please use the back side of the exam as scratch paper.
• In case you find a question ambiguous, please write down your assumption and answer the question
accordingly.

• You may use temporary relations, if you like, for any of the queries below. If you use the same
temporary relation for a second exam question, you must redefine the relation in your answer to the
second question. In other words, your answer to each question should be self-contained.

Problem 1 2 3 4 5 Total

Points 10 20 15 20 35 100

Score

Grader

Turn over the page when instructed to do so. Good Luck!

1 of 11
NetID: CS411 Midterm Exam

I. [T rue/False questions, 10 points] For each of the following statements, indicate whether it is TRUE
or FALSE by circling your choice. You will get 1 point for each correct answer, -1 point for each incorrect
answer, and 0 point for each answer left blank.

1. True False
Any candidate key of a relation is a super key for that relation.
2. True False
All attributes of a relation together form a super key for the relation.
3. True False
A relation decomposition is not always lossless.
4. True False
One weakness of triggers is that they can only be activated for insertion and update operations.
5. True False
The null value approach in translating subclass relationship in the ER model always saves space
6. True False
The default policy for the foreign-key constraint sets the value of a foreign-key to null when the tuple
of its referenced attribute gets deleted.
7. True False
Join is one of the five basic operation in relational algebra.
8. True False
In a relation R(X, Y, Z, T ), if XY → Y Z then X → Z.
9. True False
The primary key of a relation may contain a NULL value as a value for their components.
10. True False
The expression “(0 != NULL)” in the WHERE clause of a SQL query is evaluated to be true.

2 of 11
NetID: CS411 Midterm Exam

2. ER Diagram and Relational Schema (20 points; 10 points for (a) and (b))
Your task is to design a database for a banking system, which maintains information about customers and
their accounts.

(a) Draw an ER diagram to model the application with the following assumptions:

• Each customer has a name, a permanent address, and a social security number.
• Each customer can have multiple phone numbers, and the same phone number may be shared by
multiple customers.
• A customer can own multiple accounts, but each account is owned by a single customer.
• Each account has an account number, a type (such as saving, checking, etc), and a balance.
• The bank issues an account statement for each account and mails it to its account owner every
month. As time goes on, there will be multiple statements of the same account.
• Each statement has an issued date and a statement ID. All the statements of the same account
have different statement IDs, but two different accounts could have statements with the same
statement ID. For example, it is possible that account A has a statement with ID ‘123’, while
account B has another statement with the same ID ‘123’.

If you need additional assumptions to complete your diagram, please state your assumptions here as
well.
Solution:
The solution is one of the many possible solutions. See Figure 1.
Points breakdown:
• 2 points for reasonable entity sets
• 2 points for reasonable keys
• 2 points for correct relations
• 2 points for correct notations
• 2 points for recognizing statements is a weak entity (depending on your assumptions, phone may
be a weak entity as well)

3 of 11
NetID: CS411 Midterm Exam

Figure 1: ER Diagram

4 of 11
NetID: CS411 Midterm Exam

(b) Translate your ER diagram into a set of relations. Select approaches that yield the fewest number of
relations; merge relations where appropriate.

Solution:
Customers(ssn, name, permAddress)
Accounts(number, type, balance, ssn)
Has(ssn, number)
Statements(number, ID, date)
Points breakdown:
• 4 points for translating entity sets and relations correctly

• 4 points for merging various relations correctly and appropriately (ie. merging of many-to-many rela-
tions is bad)
• 2 points for correct key identification

• part b) is graded based on part a) answer. No points are taken off for propagation errors from part a).

5 of 11
NetID: CS411 Midterm Exam

3. Functional Dependencies and Normal Forms (15 points; 5 points for (a), 2 points for (b)
and (c), and 6 points for (d))

Consider a relation R(A, B, C, D, E) with FD’s AB → C, CD → E, C → A, C → D, D → B.

(a) Determine all the keys of relation R. Do not list superkeys that are not a key.
Solution:
Keys: AB, AD, C
To get the key AB, we can do the following:
• From AB → C and C → D, we obtained AB → D.
• From AB → C and AB → D, we obtained AB → CD.
• From AB → CD and CD → E, we obtained AB → E.
To get the key C, we can do the following:
• From C → A and C → B, we obtained C → AB.
• From AB, we can obtain the rest of the attributes.

To get the key AD, we can do the following:

• From D → B, we can get AD → AB.
• From AB, we can obtain the rest of the attributes.
Points breakdown:

• 3 points for correct keys.

• 2 points for work shown.
(b) List all the FD’s that violate 3NF, if any.
Solution: No violations.
Points breakdown:
• no points are taken off for propagation errors.
(c) List all the FD’s that violate BCNF.
Solution: Violation: D → B
Points breakdown:
• no points are taken off for propagation errors.

6 of 11
NetID: CS411 Midterm Exam

(d) Decompose R using the BCNF decomposition algorithm. Show your work and summarize your final
set of relations.
Solution:
Use the violation D → B to decompose the relation.
R1=DB, R2=ACDE
Points breakdown:

• 4 points for correct decomposition

• 2 points for work shown.
• no points are taken off for propagation errors.

7 of 11
NetID: CS411 Midterm Exam

4. Relational Algebra (20 points; 10 points for (a) and (b))

Tables 1, 2, and 3 on the next page show an example instance of corporate database for a coalition of retail
stores. Those tables maintain information about stores, products, and the inventories of products in different
stores. We assume that all the stores sell each product at the same price in Table 2.
Write down queries in relational algebra for the following questions. Do not use cartesian products if
you can use joins instead. Please refer to the Store relation, Product relation, and Inventory relation as S,
P , and I respectively in your solutions. We here assume that relational algebra supports the set semantics;
that is, each operation in relational algebra eliminates duplicate tuples from the output relation.

(a) Return the name of the stores that have at least one product in their inventories whose unit price is
greater than 2 USD.
πStoreN ame σU nitP rice>2 ( S ./ I ./ P)

(b) Return the name of the products with the maximum unit price.
We first find out the products that do not have the maximum price. Let’s rename Product relation to
P1 and P2 :

A = πP roductN ame (P1 ./(P1 .P roductN ame!=P 2.P roductN ame)and(P1 .U nitP rice<P2 .U nitP rie) P2 )

Then we get the answer by computing the difference between the complete list of the product names
and the above relation.

Answer = πP roductN ame P1 - A.

8 of 11
NetID: CS411 Midterm Exam

Table 1: Store relation: S

StoreID StoreName
1 Kmart
2 Walmart
3 Safeway
4 Meijer
5 Schnucks
6 Xmart
7 Ymart

Table 2: Product relation: P

ProductName UnitPrice(in USD)
Bread 1.6
Cheese-Cheddar 2
Cheese-Feta 4
Cheese-Livarot 9
Cheese-Mozzarella 5
Cucumber 4
Lettuce 1
Onion 2
Potato 2.5
Tomato 2.1
Tuna 0.5
Yam 4

Table 3: Inventory relation: I

ProductName StoreID
Bread 2
Cheese-Cheddar 4
Cheese-Cheddar 5
Cheese-Feta 7
Cheese-Livarot 4
Cucumber 1
Lettuce 4
Onion 1
Potato 5
Tomato 3
Tuna 5
Yam 3

9 of 11
NetID: CS411 Midterm Exam

5. SQL (35 points; 5 points for (a), 10 points for (b), (c), and (d))
Consider the relations given in problem 4. Write down the following queries in SQL. You do not have to
worry about duplicate tuples in the output relations of your queries.
(a) Return the names of the stores that have Tomato and Lettuce in their inventories.

(Select StoreName
From Store,Inventory
Where Store.StoreID = Inventory.StoreID and Inventory.Product =’Tomato’)
INTERSECT
(Select StoreName
From Store,Inventory
Where Store.StoreID = Inventory.StoreID and Inventory.Product =’Lettuce’)

(b) Return the names of the products whose unit prices are below the average unit price of all the products.

Select ProductName
From Product
Where UnitPrice <
( Select Avg(UnitPrice)
From Product)

10 of 11
NetID: CS411 Midterm Exam

(c) Each store has an average unit price. It is the average of the unit prices of the products available at
the store. Return the average unit prices of all stores. Your query should produce pairs of store names
and their average unit prices.

Select StoreName, Avg(UnitPrice)

From Store, Inventory,Product
Where Store.StoreID = Inventory.StoreID and Inventory.ProductName = Product.ProductName
GroupBy StoreName

(d) Return the product(s) that are available in at least three stores.

Select ProductName
From Inventory
GroupBy ProductName
Having (Count(distinct StoreID) >= 3)

11 of 11
 NetID:

CS411 Database Systems

Spring 2010, Prof. Chang
Department of Computer Science
University of Illinois at Urbana-Champaign

Midterm Examination
March 2, 2010
Time Limit: 75 minutes

• Print your name and NetID below. In addition, print your NetID in the upper right
corner of every page.

Name: NetID:

• Including this cover page, this exam booklet contains 9 pages. Check if you have
missing pages.
• The exam is closed book and closed notes. No calculators or other electronic devices
are permitted. Any form of cheating on the examination will result in a zero grade.
• Please write your solutions in the spaces provided on the exam. You may use the blank
areas and backs of the exam pages for scratch work. Please do not use any additional
scratch paper.
• Please make your answers clear and succinct; you will lose credit for verbose, convo-
luted, or confusing answers. Simplicity does count!
• Generally, we think one minute per point is a reasonable allocation of time; so plan
your time accordingly. You should look through the entire exam before getting started,
to plan your strategy.

Problem 1 2 3 4 Total

Points 36 28 18 18 100

Score

Grader

1
 NetID:

Problem 1 (36 points) Misc. Concepts

For each of the following statements:

• for true/false choices, indicate whether it is TRUE or FALSE by circling your choice, and
provide an explanation to justify;

• for short answer questions, provide a brief answer with clear explanation.

You will get 3point for each correct answer with correct explanations, and no penalty (of negative
points) for wrong answers.
Notice: To get full credit, you need to provide both correct answer, and correct explanations. If
you simply choose True or False without explaining, you will automatically lose 1 point for the
question. Moreover, since the questions here are true/false choices, or short answer questions, no
partial credit is given for incomplete answers.

(1) False

An E-R diagram with m entities and n relationships will translate to m+n tables.
⇒ Explain: Many-to-one relationships can often be merged and therefore reduce the overall
number of tables needed.

(2) True

A table with two attributes, such as R(A, B), must be in BCNF.

⇒ Explain: The only two non-trivial FDs for this relation will be A → B or B → A, and
both does not violate BCNF.

(3) True or False

An SQL query can often be translated into multiple relational algebra expressions.
⇒ Explain: The answer to this question can be true or false based on different assumptions.
Our grading takes your assumption into consideration, and grades accordingly.

(4) Give a relation that is in 3NF but not in BCNF.

⇒ Explain: A relation R(A,B,C) with two FDs: AB → C and C → A. The second FD
violates BCNF, since C is not a superkey. However, it does not violate 3NF as A is part of a
key (AB).

(5) True
In relational algebra, join is a derived operator.
⇒ Explain: It can be derived from Cartesian product and selection.

2
 NetID:

(6) False
Schema normalization often contributes to enhancement of query processing efficiency.
⇒ Explain: Schema normalization is mainly for reducing redundancy, rather than for improv-
ing query efficiency. In fact, it often introduces more query processing, as decomposed tables
have to be joined online for answering queries.

(7) True
For relation R(A, B, C), if A is a key, then the decomposition into R(A, B) and S(A, C) is
lossless.
⇒ Explain: You can always reconstruct the original table (A,B,C) by joining (A,B) with
(A,C), since A is a key. (many of your wrong answers give examples where A is not a key)

(8) True
In the following relation R(A, B, C, D), the F.D. AC → B is not violated.

A B C D
==============================
1 3 2 2
2 3 2 4
3 1 3 6
3 1 1 12

⇒ Explain: In the 4 tuples given, AC → B is not violated. You will also get full credit, if you
choose False and argue that FD dependence can not be inferred from a set of given tuples
(and should rather be derived based on the property of the relation).

(9) In the above relation R(A, B, C, D), what are possible keys of the table?
⇒ Explain: Possible keys are D, or AC. You will also get full credit, if you argue we can not
inferred keys from the set of 4 tuples.

(10) Provide a lossless-join decomposition of the above relation R(A, B, C, D) (into to two rela-
tions). If there is no such decomposition, explain why.
⇒ Explain: You can decompose it into (ABC) and (ACD), just based on the 4 tuples given in
the table. You will also get full credit, if you argue no such decomposition is possible, since
we do not know the FDs for the relation.

3
 NetID:

(11) For the above relation R(A, B, C, D), write a relational algebra expression to return the lowest
value of D.
⇒ Explain: πD R − πR1.D (R1⋊
⋉R1.D>R2.D R2) where R1 and R2 are just renames of R.

(12) What is the result of the following query, for the above relation R(A, B, C, D)?

select A, B
from R
where C >
(select D from R where A = 3)

⇒ Explain: This query will report error. The reason is the subquery returns a scalar, and
therefore can not be compared against a single data value. You would need to add “ANY”
or “ALL” before the subquery for the query to run.

4
 NetID:

Problem 2 (28 points) Database Design

You have been asked to design a database for the university administration, which records the
following information:

1. All students necessarily have a unique student ID, a name, and a university email address.
Each student is also either an undergraduate or a graduate student.

2. Each graduate student has an advisor.

3. Each undergraduate student has a major.

4. Students take courses. A student may take one course, multiple courses, or no courses.

5. Each course has a course number, course name, and days of the week the course is scheduled.

6. Each course has exactly one head TA, who is a graduate student.

7. Every head TA has an office where he or she holds office hours.

(a) Draw an ER diagram for this application. Be sure to mark the multiplicity of each relationship
of the diagram. Decide the key attributes and identify them on the diagram. Please state all
assumptions you make in your answers. (16 points)

Figure 1: ER Diagram

5
 NetID:

(b) Translate your ER diagram into a relational schema. Select approaches that yield the fewest
number of relations; merge relations where appropriate. Specify the key of each relation in your
schema. (12 points)

Undergrad(studentID, name, email, major)

Grad(studentID, name, email, advisor)
TakeCourse(studentID, course number)
Course(course number, studentID, course name, days of week, office)

6
 NetID:

Problem 3 (18 points) Relational Algebra

Nowadays, web search engine is widely used by people all over the world to find useful information.
To analyze users’ search behaviors, a query log stores the history of users’ queries and clicked URLs.
There are two relations in a query log:

• Request(T ime, U serID, Query, Results), which lists the time, the userID, the query he/she
requested, the number of search results for one search process.

• Click(T ime, U serID, Query, U RL), which lists the time, the userID, the query he/she re-
quested and the URL clicked on after request this query.

For both Request and Click relation, the attributes (T ime, U serID) forms the only key which
implies one user can request only one query or click on only one URL each time.
Example instances of these two relations are given here:

Request Relation
Time UserID Query Results
2010-02-12 19:00:00 U001 NBC 100
2010-02-12 19:01:00 U001 NBC Olympics 50
2010-02-12 18:54:00 U002 NBC 2010 80
2010-02-13 09:00:05 U001 Olympics wiki 70
2010-02-15 19:27:08 U003 Olympics 2010 medals 5
2010-02-16 13:24:45 U004 NBC Olympics 50
2010-02-16 13:25:55 U004 Vancouver 2010 95
2010-02-16 17:11:56 U002 NBC 2010 80
2010-02-20 20:13:45 U005 Olympics ski 54
2010-02-20 20:45:34 U005 Olympics Austria medals 7

Click Relation
QueryID UserID Query URL
2010-02-12 19:01:06 U001 NBC Olympics www.nbcolympics.com
2010-02-12 19:01:34 U001 NBC Olympics www.nbcolympics.com/video
2010-02-12 18:54:01 U002 NBC 2010 www.nbcolympics.com
2010-02-15 19:27:22 U003 Olympics 2010 medals www.vancouver2010.com
2010-02-15 19:29:01 U003 Olympics 2010 medals www.vancouver2010.com/olympic-medals/
2010-02-16 13:25:58 U004 Vancouver 2010 www.vancouver2010.com
2010-02-16 17:12:56 U002 NBC 2010 www.nbcolympics.com
2010-02-20 20:14:00 U005 Olympics ski www.usskiteam.com
2010-02-20 20:45:40 U005 Olympics Austria medals en.wikipedia.org/wiki/Austria at the Olympics
2010-02-20 20:45:50 U005 Olympics Austria medals www.vancouver2010.com

Note that not all the queries have corresponding clicked URLs since a user might not be interested
in any returned URL. Sometimes user may click on multiple URLs using one query.
Though different from the actual case, you can assume that the number of returned search results
associated with the same query doesn’t change no matter when and who request this query.

7
 NetID:

Answer the following queries using relational algebra. Your answer should work for any instance of
the database, not just this one.

(a) List all the queries that do not have any corresponding URL. (8 points)
Solution:
πQuery Request − πQuery Click
(b) List all the queries that have been requested by different users. (10 points)
Solution:
πRequest1 .Query (Request1 ⊲⊳C Request2 )
where, Request1 and Request2 are both type of Request
C = (Request1 .Query = Request2 .Query AND Request1 .U serID <> Request2 .U serID)

8
 NetID:

Problem 4 (18 points) SQL

Consider the same schema as Problem 3. Answer the following queries using SQL. Your statements
should work for any instance of the database, not just this one. You do not need to remove dupli-
cations in your results.

(a) List all the users (userID) who either requested the query “NBC 2010” or clicked on the URL
“www.nbcolympics.com.” (8 points)
Solution:
select UserID from Request where Query = “NBC 2010”
union
select UserID from Click where URL = “www.nbcolympics.com”

(b) Return the query whose corresponding URLs contain “www.vancouver2010.com” and, among
these queries, has the minimum number of returned search results. (10 points)
Solution:
select Request.Query from Request, Click
where Request.Query = Click.Query
and Click.URL = “www.vancouver2010.com”
and Results <= All
(select Results from Request, Click
where Request.Query = Click.Query
and Click.URL = “www.vancouver2010.com”)

9
Name:

CSE 444, Fall 2010, Midterm Examination

10 November 2010

Rules:

• Open books and open notes.

• No laptops or other mobile devices.
• Please write clearly.
• Relax! You are here to learn.

Question Max Grade

1 30
2 30
3 40
Total 100

1
Initials:

1. (30 points) SQL

Consider a database with the following three relations.

Sensor( sid, type, location, description )

Temperature ( sid, time, tempvalue )
Pressure ( sid, time, pressvalue )

Temperature.sid is a foreign key that references Sensor.sid.

Pressure.sid is a foreign key that references Sensor.sid.

This database holds a set of sensor readings (temperature and pressure) recorded by a set of sensors
deployed at various locations. Some sensors record only temperature readings. Others record only
pressure readings. Some record both. Each sensor produces a reading every minute (pressure reading,
temperature reading, or both). Each sensor is associated with a given type, location, and description.

(a) (8 points) Write a SQL query that returns the type and description of all sensors that have
produced both temperature readings and pressure readings. Each unique combination of type
and description should appear only once in the output.

Solution:

SELECT DISTINCT S.type, S.description

FROM Temperature T, Pressure P, Sensor S
WHERE T.sid = S.sid AND P.sid = S.sid

2
 Initials:

(b) (8 points) Write a SQL query that computes the highest ever pressure reading among all sensors
that recorded both pressure and temperature readings.

Solution:

SELECT MAX(PI.pressvalue)
FROM Pressure PI, Sensor SI, Temperature TI
WHERE PI.sid = SI.sid AND TI.sid = SI.sid

3
 Initials:

(c) (14 points) Consider the set of sensors that record BOTH temperature and pressure (your query
will have to compute that set). Write a SQL query that computes, for these sensors only, the sid
and average tempvalue for all sensors that recorded the highest ever pressure reading. Note that
more than one sensor could have recorded the same, highest pressure reading.

Solution:
We can use our query from the previous question in a subquery to find the sid and average
tempvalue for all sensors that produced such a high pressure reading:
SELECT S.sid, AVG(T.tempvalue)
FROM Pressure P, Sensor S, Temperature T
WHERE P.sid = S.sid AND S.sid = T.sid
AND P.pressvalue >= (SELECT MAX(PI.pressvalue)
FROM Pressure PI, Sensor SI, Temperature TI
WHERE PI.sid = SI.sid AND TI.sid = SI.sid)
GROUP BY S.sid

Alternate solution:
SELECT S.sid, AVG(T.tempvalue)
FROM Pressure P, Sensor S, Temperature T
WHERE P.sid = S.sid AND S.sid = T.sid
AND P.pressvalue >= ALL (SELECT PI.pressvalue
FROM Pressure PI, Sensor SI, Temperature TI
WHERE PI.sid = SI.sid AND TI.sid = SI.sid)
GROUP BY S.sid

4
 Initials:

2. (30 points) Conceptual Design

(a) (12 points) Draw an E/R diagram describing the following domain:
• Users have a single attribute uid (key).
• AnonymousUsers are a type of Users with attribute country.
• NamedUsers are a type of Users with attributes name and email.
• Blogs have attributes bid(key), title, and topic.
• A NamedUser can be a friend-of zero or more other NamedUsers. Each friend-of rela-
tionship has an associated start date.
• A NamedUser can own many Blogs, but each blog is owned by exactly one user.
• A User can post to zero or more Blogs. Each post has a given content. Many users can
post to a blog.
Your answer should consist of an E/R diagram, which includes entity sets, attributes, relationships,
ISA relations. Indicate the type of each relationship with appropriate arrows (one-one, one-many,
etc.).

Solution:

uid

date
content Users

Friend
isa isa ‐of

post Anonymous Named

country
Users Users name

own
email

Blogs

bid 6tle topic

5
 Initials:

(b) (8 points) Consider the following relational schema and set of functional dependencies. (a)
List all superkey(s) for this relation. (b) Which of these superkeys form a key (i.e., a minimal
superkey) for this relation? Justify your answer in terms of functional dependencies and closures.

R(A,B,C,D,E) with functional dependencies CD → E and A → B.

Solution:
(a) A superkey is a set of attributes X s.t. X + = all attributes.

From the FDs above, we can derive:

{A, B, C, D, E}+ = {A, B, C, D}+ = {A, C, D, E}+ = {A, C, D}+ = {A, B, C, D, E}

Hence,
{A, B, C, D, E}, {A, B, C, D}, {A, C, D, E}, and {A, C, D} are all superkeys.

(b) A key is a set of attributes which form a superkey and for which no subset is a superkey. In
our example, {A, C, D} is the only key.

6
 Initials:

(c) (10 points) Decompose R into BCNF. Show your work for partial credit. Your answer should
consist of a list of table names and attributes and an indication of the keys in each table (underlined
attributes).

Solution:
Both functional dependencies violate BCNF.
Try {A}+ = {A, B}. Decompose into R1(A,B) and R2(A,C,D,E).

R1 has two attributes, so it is necessarily in BCNF.

R2 is not in BCNF, since {C, D} is not a key and we have CD → E.

Try {C, D}+ = {C, D, E}. Decompose into R3(C, D, E) and R4 (A, C, D )

End result: R1(A,B), R3(C, D, E), and R4 (A, C, D )

7
 Initials:

3. (40 points) Transactions

(a) (6 points) In the ARIES method, assuming NO checkpoints have been used, explain what
happens during the ANALYSIS pass of recovery. Your answer should indicate at least (1) what
part of the log the system reads and in what direction and (2) what data structures the system
rebuilds.

Solution:
In the absence of checkpoints, the analysis pass reads the log from the beginning to the end. It
rebuilds the Dirty Pages Table and the Transactions Table to determine the state of the system
as of the time of the crash. It rebuilds these data structures by updating them according to the
log records that it encounters during the forward scan.

(b) (6 points) After the analysis pass, the protocol performs a REDO pass. Explain (1) where does
REDO start reading the log and in what direction it reads the log, (2) what happens during the
REDO pass.

Solution:
The REDO pass begins at the log record whose LSN corresponds to the earliest recoveryLSN of
all the entries in the Dirty Page Table. From that point, the REDO pass redoes the updates of
all transactions (committed or otherwise). At the end of this pass, the database is in the same
state as it was right before the crash.

8
 Initials:

(c) (6 points) The last pass during recovery is the UNDO pass. Explain (1) where does the UNDO
start reading the log and in what direction it reads the log, (2) what happens during the UNDO
pass.

Solution:
The UNDO pass scans backward from the end of the log. The pass undoes the updates by all
transactions that had not committed by the time of the crash. These transactions can be found in
the Transactions Table rebuilt during the analysis pass. All updates are undone unconditionally
(since the REDO pass ensured that all logged updates have been applied to affected pages).
For each update that is undone, the undo operation is logged with a Compensation Log Record
(CLR).

9
 Initials:

(d) (7 points) Consider the following two transactions and schedule (time goes from top to bottom).
Is this schedule conflict-serializable? Explain why or why not.

Transaction T0 Transaction T1
r0 [A]
w0 [A]
r1 [A]
r1 [B]
c1
r0 [B]
w0 [B]
c0

Solution:
The schedule is not conflict serializable because the precedence graph contains a cycle. The graph
has an edge T0 → T1 because the schedule contains w0 [A] → r1 [A]. The graph has an edge
T1 → T0 because the schedule contains r1 [B] → wo [B].

Because of W0[A] R1[A]

T0 T1

Because of : R1[B] W0[B]

10
 Initials:

(e) (7 points) Show how 2PL can ensure a conflict-serializable schedule for the same transactions
above. Use the notation Li [A] to indicate that transaction i acquires the lock on element A and
Ui [A] to indicate that transaction i releases its lock on A.

Solution:

Multiple solutions are possible.

Transaction T0 Transaction T1
L0 [A]
r0 [A]
w0 [A]
L1 [A] → blocks
L0 [B]
r0 [B]
w0 [B]
U0 [A]
U0 [B]
c0
L1 [A] → granted
r1 [A]
L1 [B]
r1 [B]
U1 [A]
U1 [B]
c1

11
 Initials:

(f) (8 points) Show how the use of locks without 2PL can lead to a schedule that is NOT conflict
serializable.

Solution:

Transaction T0 Transaction T1
L0 [A]
r0 [A]
w0 [A]
U0 [A]
L1 [A]
r1 [A]
U1 [A]
L1 [B]
r1 [B]
U1 [B]
c1
L0 [B]
r0 [B]
w0 [B]
U0 [B]
c0

12
Problem 1: ER Diagram Design
The company you work for wants to digitize their time cards. You have been asked to design the
database for submitting and approving time cards. Draw the database ER diagram with the following
information:

 A timecard should have hours worked and date submitted

 Each timecard is associated with exactly one employee
 Each timecard should have a unique id
 Each timecard has a status: it is either approved, not approved, or pending
 Each employee has a unique id
 Each employee has a name and address.
 Each employee submits a time card every pay period. i.e. In 1 year, they will submit multiple time
cards
 Each employee either has direct deposit or physical check as their method of payment
 Each employee is associated with exactly one manager
 Each manager has a unique id and a name
 Each manager is in charge of multiple employees
 Each manager approves time cards for multiple employees

If you feel that you must make some assumptions, please state them clearly so that they are easily
understood by the graders. Remember to indicate the key for each entity, as well as the multiplicity of
each relationship (e.g. one-to-many) using the appropriate notation.

Solution:
Use the above diagram for the following
questions!
Problem 2: More ER Fun
Answer the following questions based on the original ER Diagram in Figure 1:

1. Could the "Date of Measure" attribute of the Measurement entity be the key for the entity, instead
of the "MeasureID"? Why or Why Not?
2. We want to add a "Tools Used" attribute, which will store the tool(s) used to make measurements,
but are not sure where this attribute belongs. We do know that one person might use multiple
tools for different trees, and also that more than one tool may be used to measure the same tree
(for example, different tools may be necessary to measure the same tree in the summer, than in
the winter.) Where could we add this attribute? Choose one or more answers from the 3 entities
and 2 relationships (we would add the attribute only once, but if you believe there are multiple
possible places to add the attribute, we ask you to identify all potential candidates):
a. the entity Person
b. the relationship Has
c. the entity Measurement
d. the relationship Takes
e. the entity Tree
3. There are two roles that people can have, a Ranger and a Volunteer. Using what you know of
subclassing, add these as two new entities in the ER Diagram. Add a reasonable attribute to each
of the new entities. Just redraw the relevant part of the diagram that needs to change. Can we
eliminate the role attribute fromPerson?

Solution
1. It would not be a good key because there may be many measurements made on the same day.
Therefore this attribute is not unique, and would not make a good key.
2. The relationship Has, the entity Measurement, and the relationship Takes are all correct answers
(b, c, d). The entity Person and the entity Tree are not (a, e).
3. See Diagram. Any appropriate attributes to the subclasses can be accepted. You can eliminate
the role type attribute. However, we may not want to, as keeping the payment type attribute may
facilitate some queries, and may make creating new payment types easier.
Problem 3: ER Diagram Translation
Use the original ER Diagram in Figure 1 for this problem (without any changes you may have made
during your work in problem 2).

1. Translate the ER diagram to a relational design. Try to minimize the number of relations your
solution has, and merge relations where appropriate. Don't forget to specify the keys.
2. Assume now that we have changed the ER diagram so that the Takes relationship is now many-
many, with the interpretation that it might take several people to determine a single measurement.
How will this change the relational design?

Solution:

1. Person(UserID, Name, Role, Address), Tree(TreeID, Latitude, Longitude, Species),

Measurement(MeasureID, Date_of_Measure, Height, Diameter, TreeID, UserID) (because
Measurements had a 1-many relationship with both Person and Tree, the Measurement table can
store those relationships).
2. Now the many-many relationship needs its own table, Takes(UserID, MeasureID) and
the UserID column would be removed from the Measurement table. 6 points
Problem 4: SQL Statements
Grading: 25 points

Use the original ER Diagram in Figure 1 for this problem (without any changes you may have made
during your work in problems 2 or 3).

1. Write the SQL command that would define each table. Your definition must include correct data
types with correct sizes for each field, key and unique declarations, and NULL constraints for
fields. Please assume the following data types and lengths:
a. the three ID attributes are all of type char, length 10
b. Latitude and Longitude are of type decimal, precision value 10, scale value 6
c. Height and Diameter precision value 10, scale value 6
d. DateOfMeasure is of type date
e. the rest of the attributes are all of type varchar, length 50
2. Write the SQL Command to insert one tuple of data into each table. The data you insert can be of
your choosing, but must adhere to the data types and constraints of the tables.
3. We have decided that the Address property is no longer relevant. We do however wish to be able
to email staff members. Change the Person table to remove the Address field and add an Email
field.
4. Write a SQL Statement to add an Email address to the person you inserted into the Person table,
now that the table has an Email field and not an Address field.

Solution:

1. CREATE TABLE Person(UserID char(10) PRIMARY KEY, Name varchar(50) NOT NULL, Address
varchar(50) NOT NULL, Role varchar(50));
CREATE TABLE Tree(ID char(10) PRIMARY KEY, Species varchar(50) NOT NULL, Latitude
Decimal(10,6) NOT NULL, Longitude Decimal(10,6) NOT NULL);
CREATE TABLE Measurement(MeasureID char(10) PRIMARY KEY, DateOfMeasure date NOT NULL,
Height Decimal(10,2) NOT NULL, Diameter Decimal(10,2) NOT NULL);
9 points, 3 points per Create statement. Only the primary keys need to be specified as NOT NULL, the
rest are ok either way

2. INSERT INTO Person (UserID, Name, Address, Role) VALUES ('12345', 'John Smith', '123 Science
Lane, Urbana, IL', 'Ranger');
INSERT INTO Tree(ID, Species, Latitude, Longiture) VALUES ('987', 'Fir', 9.353234, 39.234562);
INSERT INTO Measurement(MeasureID, DateOfMeasure, Height, Diameter) VALUES ('432', '2011-01-
30', 15.50, 3.10);
6 points, 2 points per Insert Statement

3. ALTER TABLE Person DROP Address;

ALTER TABLE Person ADD Email varchar(50);
5 points

4. UPDATE Person SET Email = '[email protected]' WHERE UserID = '12345';

 CS411 Tutorial 1 1

Problem 1: ER Diagram
You have been hired to develop an account management system for a company. The company
sells packages that allow consumers to download books in either text or audio format. Draw the
database ER diagram with the following information:

 A package allows the consumer to download a given number of books or audiobooks a month, depending
on the package type.
 Each package has a unique id.
 Each package also must have a type, noting what kind of downloads it provides: either books or
audiobooks but not both.
 Each package also must have a tier, representing the number of monthly downloads allowed by this
package (e.g. “Tier10” means 10 monthly downloads are allowed, "Tier50" means 50 monthly downloads
are allowed, etc.)
 Each account has a unique id.
 Each account has one or more packages.
 Each account is associated with one or more credit cards
 Each credit card is associated with exactly one account.
 Each credit card must have a short nickname to identify it to its account (e.g. “card1”, “visa”), a credit card
number, and an expiration date
 Each account is owned by exactly one customer
 Each customer must have a customer id, a name, and one contact email, and may also specify one contact
phone number.
 CS411 Tutorial 1 2

Problem 2: ER Diagram continued

Answer the following questions based on your diagram from Problem 1:

1. Would expiration date be a good key for the Credit Card entity? Why or why not?
2. For the Package entity, which attributes must not be null? What about for the Customer
entity?
3. There are two types of packages: book and audiobook. Using your knowledge of
subclassing, introduce these as two new entities in the ER diagram. Add a reasonable
attribute to each of the new entities. Can we eliminate the type attribute of the Package
entity? Just redraw the relevant part of the diagram that needs to change.
 CS411 Tutorial 1 3

Problem 3: ER Diagram translation

Answer the following questions based on your original diagram from Problem 1:

1. Translate the ER diagram from Problem 1 to a relational design. Try to minimize the
number of relations your solution has, and merge relations where appropriate. Don‟t
forget to specify the keys.
2. Translate the ER diagram from Problem 2 question 3 (only the Package entity and the
two new entities) to a relational design using OO, ER, and NULL value approaches.
Don‟t forget to specify the keys.
 CS411 Tutorial 1 4

Problem 4: SQL commands - table and field

definitions
Answer the following questions based on your relational design from Problem 3. The response to
each question should be a valid SQL command:

1. Write the SQL command that would define each table. Your definition must include
appropriate data types with reasonable sizes for each field, key and unique declarations,
and NULL constraints for fields:
1. Customer
2. Account
3. Credit Card
4. Package
2. Insert one tuple into the Customer table. You can choose arbitrary values for the data
fields of your tuple but they must conform to the table‟s defined constraints.
3. Rename the nickname field in the Credit Card table to cardID.
4. Change the type field in the Package table to have a default value of book, represented by
the value „B‟.
Homework 1 Solution
Problem 1: ER Diagram
Solution

Problem 2: ER Diagram continued

Solution
1. No - multiple credit cards may have the same expiration date.

2. Package: id, type, tier. Customer: id, name, email.

3. See diagram. Yes, we could eliminate the type attribute. However, we may not want to, as
keeping the type attribute may facilitate some queries, and may make creating new types easier.
Reasonable attributes for book: title, author, genre, number of pages, etc; for audiobook: title,
author, genre, length, etc. Any attribute that makes sense is acceptable.

Problem 3: ER Diagram translation

Solution
1. Relational design: Package(id, type, tier), Card(nickname, number, exp, account_id),
Account(id, customer_id), Customer(id, name, email, phone), HasPackage(account_id,
package_id)

2. If you chose to remove the type attribute:

 OO: Package(id, tier), PackageBook(id, tier, title), PackageAudioBook(id, tier, length)

 ER: Package(id, tier), PackageBook(id, title), PackageAudioBook(id, length)

 Null: Package(id, tier, title, length)

If you chose not to remove the type attribute:

 OO: Package(id, type, tier), PackageBook(id, type, tier, title), PackageAudioBook(id,

type, tier, length)

 ER: Package(id, type, tier), PackageBook(id, title), PackageAudioBook(id, length)

 Null: Package(id, type, tier, title, length)

(title and length were the attributes chosen for book and audiobook for this solution. If you chose
different attributes, just replace title and/or length in the answers with the attributes you chose)

Problem 4: SQL commands - table and field

definitions
Solution
1.
1. Create table Customer(id char(10) Primary Key, name varchar(50) not null, email
varchar(50) not null, phone char(10));
2. Create table Account(id char(10) Primary Key);
3. Create table CreditCard(number char(16) Primary Key, nickname varchar(20) not
null, expDate date not null);
4. Create table Package(id char(10) Primary Key, type char(1) not null, tier
varchar(10) not null);
2. Insert Into Customer(id,name,email,phone) Values('1234567890', 'John Doe',
'[email protected]', '9876543210');
3. Alter table CreditCard Drop nickname; Alter table CreditCard Add cardID varchar(20);
4. Alter table Package Drop type; Alter table Package Add type char(1) Default 'B';
 Problem 1
_
phone name number
address
number of units space
number isCovered

Property manages
Apartment has
Management
Company Parking space

supplies
number
has of staff

Leasing Office
Electricity
Provider
business
registration phone
number name number
address

Problem 2
number of number of
unit bedroom bathroom
number
address number
of units
area
has Unit
Social
manages
Sec #
Apartment
name
To property
For
management Start rate
company date
supplies
has
Tenant DOB
To Electricity
provider Lease
has
Contract
To parking (short term) by
space
contract ID
To leasing
office end
Long term Lease date
Contract
Problem 3
a) Convert the E/R diagram of Problem 1 to a relational database schema. Donʼt
forget to indicate the keys (by underlining them) for each relation. Note that you
donʼt need to combine the relations for removing duplications in this problem. (4
points)
ANSWER:
Property Management Company(name, phone number)
Apartment(address, company name, electricity provider reg num, number
of units, leasing office phone number, leasing office number of staff)
Parking space(apartment address, space number, isCovered)
Electricity Provider(Business reg num, name, address)

b) Convert the Leasing contract, long term contract, and monthly contract (Note
you may use different names other than ones given here, as long as you make it
clear for graders to understand) relationships in Problem 2 into relational
schema, using the following three approaches: 1. ER approach, 2. Object
oriented approach, 3. and Null approach. (6 points, 2 points for each approach)
ANSWER:
ER approach:
Short term lease contract(contract ID, rate, start date)
Long term lease contract(contract ID, end date)

OO approach:
Short term lease contract(contract ID, rate, start date)
Long term lease contract(contract ID, rate, start date, end date)

Null approach:
Lease contract(contract ID, rate, start date, end date)

Problem 4 Basic Concepts of Keys (25 points, part b 10 and other parts 5 points)
a) Given a relation R, explain how you can use the closure to test if a subset of
attributes A1, A2, ..., An is a candidate key for R.

ANSWER: By first checking that the closure of A1, A2, ..., An (i.e. { A1, A2, ...,
An}+ is all attributes of the relation R, and then checking that no subset of
A1, A2, ..., An is all attributes of R.

b) Suppose R is a relation with attributes A1, A2, ..., An. As a function of n, tell
how many superkeys R has, if
(1) the only keys are A1 and A2, and
ANSWER: Number of combinations of attributes that contains {A1}, {A2} or
{A1,A2}: 3 * 2^{n-2}
c) Show that each of the following are not valid rules about FDʼs by giving
example relations that satisfy the given FDʼs (following the ”if”) but not the FD
that allegedly follows (after the ”then”). Explain your answer and state any
assumption you make.
1. If A → B then B → A
2. If AB → C then A → C or B → C
ANSWER:
1. StudentID Surname First name
12335 Jones James
56932 Jones Bridget
Surname is functionally dependent on StudentID, but more than one
student may share a same surname.

2. SupplierID ProdID Quantity

1 4902 300
1 9088 10
2 4902 500
Multiple suppliers (of some product) may supply the same product with
varying quantity. SupplierID and ProdID are both required to uniquely
identify the quantity supplied.

d) Consider a relation with schema R(A,B,C,D) with functional dependencies

(FDʼs): BC → A, AD → B, CD → B, AC → D.
Find all the candidate keys of R.
ANSWER: BC, CD, AC

Problem 5 Functional Dependency (20 points, 5 points each)

Consider a relation with schema R(A,B,C,D) and FDs AC → B, B → A, BD → C
and D → A.
a) Describe the concept of Functional Dependency and Dependency-preserving
decomposition.
ANSWER:
Functional dependency can be thought of as a generalization of the idea of
a key for a relation. It states that if 2 tuples of a relation agrees on the
attributes A1A2…An, then they must also agree on attributes B1B2…Bn,
written A1A2…An -> B1B2…Bn. Simply put, it means attributes A1A2…An
functionally determines attributes B1B2…Bn.
Dependency-preserving decomposition is a decomposition of a relation R
into R1, R2…,Rn according to some FD where all FDs that hold on R also
hold on all “sub-relations” R1, R2…,Rn.
b) Find the closures for subsets A, BD and BC respectively.
ANSWER:
{A}+ = {A}
{BD}+ = {ABCD}
{BC}+ = {ABC}

c) List all nontrivial FDʻs that follow from the given FDʻs.
ANS: BD -> A, CD -> A, CD -> B, CD -> AB, ABD -> C, BCD -> A, ACD -> B,
BC -> A, BD -> AC

d) Find all the candidate keys for this relation (you donʻt need to list superkeys
that are not keys). Explain your answer.
ANSWER: BD, CD since {B,D} and {C,D} are both minimal set of attributes
whose closure contains all attributes {A,B,C,D,E}

Problem 6 Normalization and Schema Design (20 points, 5 points each)

Consider a relation R(A, B, C, D), with FDs AB → C , BC → D, CD → A.
a) Is R in BCNF? Explain why or why not.
ANSWER: No, R is not in BCNF since the functional dependency CD->A
violates the condition; CD is not a superkey of R.

b) We are considering to decompose R into R1 (A, B, C) and R2 (A, C, D). Is this

a lossless decomposition?
ANSWER: No.
Assume R has the following tuples to start with:
ABCD
12 3 4
14 3 2

After decomposition, R1 and R2 will have the following tuples:

R1: A B C R2: A C D
12 3 134
14 3 132
Joining R1+R2 results in Rʼ as follows:
Rʼ: A B C D
12 34
12 32
14 34
14 32
Since R is not the same as Rʼ (2nd and 3rd tuples are introduced after join),
this decomposition is not lossless.
c) Provide the BCNF normal form for this relation
ANSWER:
Decomposing using CD->A
R1(ACD), R2(BCD)

d) Provide the 3rd normal form for this relation.

ANSWER: R(ABCD) is in 3NF
Name: NetID:

MIDTERM EXAM, March 9, 2006

CS 411 Database Systems
Department of Computer Science
University of Illinois at Urbana-Champaign

Exam Rules:

1) Close book and notes, 75 minutes, scratch papers are allowed.

2) Please write down your name and NetID number NOW.

3) Please wait until being told to start reading and working on the exam.

4) No question can be asked during the exam (due to departmental regulations, to

be fair to off-campus students). If you think a problem is ambiguous, write
down your assumptions, argue that they are reasonable, then work on the
problem using those assumptions.

5) No electronic devices are permitted.

Scores:
--------------------------------------------------
Problem 1: out of 15 points
Problem 2: out of 15
Problem 3: out of 20
Problem 4: out of 30
Problem 5: out of 20
--------------------------------------------------
Total: out of 100 points

Page 1 of 6
Problem 1. (15 points) ER and translation to relation model

Translate the following ER diagram into a relational schema. For each relation in your schema, specify the
key of that relation. In translating a subclass hierarchy, use the ER style translation.

customer-id street

loan-number amount
name city

customer loan
borrows pays

account-number balance
payment

owns
account
p-number p-amount
start-date
p-date

IS-A

Savings-account Checking-account

Interest-rate Overdraft-limit

customer(customer-id, name, street, city)

loan(loan-number, amount)
payment(p-number,loan-number, p-date, p-amount)
borrows(customer-id, loan-number)
account(account-number, balance)
owns(customer-id,account-number, start-date)
Savings-account(account-number, Interest-rate)
Checking-account(account-number, Overdraft-limit)

Grading Scheme:
- 4pts for the “payment” relation
- 3pts for the “Savings-account” relation
- 3pts for the “Checking-account” relation
- 1pt for each other relation

Page 2 of 6
Problem 2. (15 points) Schema Refinement

Consider the relation R(A,B,C,D,E) with the following functional dependencies:

A → B, A → C, BC → A, D → E.

Is R in BCNF or 3NF or neither? If R is not in BCNF, decompose R into a collection of BCNF relations.
Show each step of the decomposition process.

[NOTE] FDs = {A → B, A → C, BC → A, D → E} = {A → BC, BC → A, D → E}

It is neither in BCNF nor in 3NF.

The keys are AD, BCD.
All the FDs violate the BCNF condition as none of their LHS are superkeys.

There are two possible decompositions: {ABC, DE, AD} or {BCA, BCD, DE}

Decomposition 1
Decomposing on A→BC:
R1 = (ABC), FD1 = {A→BC, BC→A}:
This is in BCNF as A and BC are keys and LHS of both FDs are super keys.
R2 = (ADE), FD2 = {D→E}:
This is not in BCNF because the key is AD where as LHS of the FDs is not a superkey.
Decomposing R2 on D→E:
R21 = (DE), FD21 = {D→E}
This is in BCNF as the key is D and LHS of the FD a superkey.
R22 = (AD), FD22 = {}
This is in BCNF as there are no FDs
Therefore the decomposed schema is:
R1 = (ABC), FD1 = {A→BC, BC→A}, R21 = (DE), FD21 = {D→E}, R22 = (AD), FD22 = {}

Decomposition 2
Decomposing on BC→A:
R1 = (BCA), FD1 = {A→BC or BC→A}
This is in BCNF as A and BC are the keys and hence LHS of all FDs are superkeys.
R2 = (BCDE), FD2 = {D→E}
This is not in BCNF as the key is BCD and hence LHS of FD D→E is not a superkey.
Decomposing R2 on D→E:
R21 = (DE), FD21 = {D→E}
This is in BCNF as D is the key and the LHS of the only FD.
R22 = (BCD), FD22 = {}
This is in BCNF because there is no FD.
Therefore the decomposed schema is:
R1 = (BCA), FD1 = {A→ BC, BC→A}, R21 = (DE), FD21 = {D→E}, R22 = (BCD), FD22 = {}

Grading Scheme:
- 5 pts for the first part and 10 pts for the decomposition.

Page 3 of 6
Problem 3. (20 points) Relational Algebra

This problem has 2 parts.

Consider the following database schema:

course (course#, dept-name)

enroll (studentID, course#, status)

(a) (10 points) Write a relational algebra expression to find the course# of all the courses offered by CS
department.

Project_{course#}(Select_{dept-name==”CS”}(Course))

Grading Scheme:
- No partial credits here.

(b) (10 points) Write a relational algebra expression to find the studentIDs of all the students who are not
enrolled in course# 411.

Project_{studentID}(enroll) – Project_{studentID}(Select_{course# == “411”}(Enroll))

Some students answered:

Project_{studentID}(Select_{course# <> “411”}(Enroll))

This answer is incorrect. Suppose there is a student who does not register for 411, however, the above query
would still return the studentID for that student if he/she is registered for any course other than 411.

And a few students answered:

Project_{studentID}(enroll – Select_{course# == “411”}(Enroll))

This was a good attempt at removing above problem, however, still suffers from that problem.

A good way to verify your relational algebra expression or sql query is to try on a small example.

Grading Scheme:
- For the two incorrect answers, we award 5/10 points.

Page 4 of 6
Problem 4. (30 points) SQL

This problem has 3 parts.

Again consider the following database schema:

course (course#, dept-name)

enroll (studentID, course#, status)

(a) (10 points) Write a SQL query that finds the studentID of all the students who are enrolled in at least one
course offered by CS department and are not enrolled in any courses offered by EE department.

{SELECT studentID
FROM course, enroll
WHERE course.course# = enroll.course# AND dept-name = ‘CS’}
EXCEPT
{SELECT studentID
FROM course, enroll
WHERE course.course# = enroll.course# AND dept-name = ‘EE’}

(b) (10 points) Write a SQL query that lists the course# of all the courses for which more than 50 students are
enrolled.

SELECT course#
FROM enroll
GROUP BY course#
HAVING count(studentID)>50

(c) (10 points) Write a SQL query that finds the studentID of all the students who are enrolled in the
maximum number of courses. (For example, suppose there are 3 students, each is enrolled in 5 courses, and
all other students are enrolled in fewer than 5 courses. Then you are expected to return the studentID of these
3 students.)

CREATE VIEW v AS
SELECT student ID, count(course#) as NumOfCourses
FROM enroll
GROUP BY studentID

SELECT studentID
FROM v
WHERE NumOfCourses=max(NumOfCourses)

Page 5 of 6
Problem 5. (20 points) Constraints and Assertions

In this problem, you are given the schema for a bank and will be required to write an assertion. The relation
Account-owners is a many-to-many relationship between the relation Customers and the relation Accounts.
The SQL definitions for these relations are as follows:

create table Customers

(customer-name char(20),
customer-address char(20),
primary key (customer-name))

create table Accounts

(account-number char(10),
balance integer,
primary key (account-number))

create table Account-owners

(customer-name char(20),
account-number char(10),
primary key (customer-name, account-number),
foreign key (customer-name) references Customers(customer-name),
foreign key (account-number) references Accounts(account-number))

Write an assertion such that for every customer at least one of the following conditions holds true:
• He (or she) is owner of at most 5 accounts
• The sum of the balance of various accounts he (or she) owns is greater than $50,000.

CREATE ASSERTION mid-term-assert AS CHECK

(NOT EXISTS (SELECT O.customer-name
FROM Accounts A, Account-Owners O
WHERE A.account-number == O.account-number
GROUP BY O.customer-name
HAVING COUNT(A.account-number) > 5
and SUM(A.balance) < 50000))

Alternative answer:
CREATE ASSERTION mid-term-assert2 AS CHECK
(NOT EXISTS (SELECT * FROM Customers C
WHERE (SELECT count(*) FROM Account-Owners O
WHERE O.customer-name == C.customer-name) > 5
AND (SELECT SUM(balance)
FROM Accounts A, Account-Owners O2
WHERE A.account-number == O2.account-number
AND A.customer-name == C.customer-name) < 50000
))
Grading Scheme:
- checking the two conditions: 8 each
- combining the two conditions: 2 points
- syntax of assertion: “CREATE ASSERTION name CHECK (some Boolean condition)”: 2 points

Page 6 of 6
Name:

CSE 444, Winter 2011, Midterm Examination

9 February 2011

Rules:
• Open books and open notes.
• No laptops or other mobile devices.

• Please write clearly.

• Relax! You are here to learn.
• An extra page is provided in case you run out of space, but make sure to put a forward reference.

Question Max Grade

1 30
2 30
3 18
4 22
Total 100

1
Initials:

1. (30 points) SQL

For a given quarter, the table Student lists all students registered at the University, the table Course
lists all courses offered, and the table Enrollment encodes the students that attend any course in the
current quarter.

Student ( sid, sname )

Course ( cid, cname )
Enrollment ( cid, sid )

Enrollement.sid is a foreign key that references Student.sid.

Enrollement.cid is a foreign key that references Course.cid.

(a) (8 points) Assume that no two courses have the same name. Call a course “big” if the number
of students enrolled is at least 50. Write a SQL query that returns the names of all big courses in
decreasing order of their enrollment size.

Solution:

SELECT cname
FROM Course C, Enrollment E
WHERE C.cid = E.cid
GROUP BY cname
HAVING COUNT(*) >= 50
ORDER BY COUNT(*) DESC

2
 Initials:

(b) (8 points) A “classmate” of a student is another student who is enrolled in at least one same
class. Write a SQL query that returns all pairs of classmates. You should return their student
IDs, as their names do not uniquely identify them.

Solution:

SELECT DISTINCT E1.sid AS sid1, E2.sid AS sid2

FROM Enrollment E1, Enrollment E2
WHERE E1.cid = E2.cid AND E1.sid != E2.sid

3
 Initials:

(c) (14 points) Write a SQL query that computes for each student id in the Student table the total
number of different classmates that this student has across all courses. Note that a student may
be so busy with research (or parties ^)¨ that he or she may choose not to enroll to any course
that quarter, and hence has no classmates. Such students should still appear in the result, with
0 as the number of their classmates.
Hint: this is not required, but it may help you to make use of an earlier query.

Solution:
We can use our query from the previous question in a subquery:
SELECT S.sid, COUNT(T.sid2)
FROM Student S LEFT OUTER JOIN (SELECT DISTINCT E1.sid AS sid1, E2.sid AS sid2
FROM Enrollment E1, Enrollment E2
WHERE E1.cid = E2.cid AND E1.sid != E2.sid) T
ON S.sid = T.sid1
GROUP BY S.sid

Slightly different alternative:

SELECT S.sid, COUNT(DISTINCT T.sid2)

FROM Student S LEFT OUTER JOIN (SELECT DISTINCT E1.sid AS sid1, E2.sid AS sid2
FROM Enrollment E1, Enrollment E2
WHERE E1.cid = E2.cid) T
ON S.sid = T.sid1 AND S.sid != T.sid2
GROUP BY S.sid

4
 Initials:

2. (30 points) Conceptual Design

(a) (12 points) A friend is impressed by your skills in CSE 444, and asks you to help her design
a database to model doctors’ offices across Seattle. She started the E/R diagram below which
already contains all the entity sets, their attributes and relationships she wants you to consider.
You need to finish the diagram and define a number of constraints to ensure that you model the
semantics of an office as closely as possible.
Make sure that you do not impose additional constraints not defined by the model.
i. (4 points) An office may be managed by at most one doctor. A doctor is uniquely identified
by their badgeId, and may manage more than one office. Draw the necessary constraints to
capture this requirement.

ii. (5 points) An office is identified by its address, and contains one or more exam rooms.
A room can be identified by its room number, and the office that it is in. Capture these
constraints by drawing on or modifying the diagram where necessary.

iii. (3 points) When a patient visits an office, he or she has a consultation with a doctor in an
examination room. Each patient is uniquely identified by their SSN.

Draw your answer directly onto the figure below. Make sure that any arrows, lines, double lines
etc, are clearly visible.

Solution:

phone#   address   badgeId   name  

Office   Manages   Doctor  

ssn  

Inside   Consults   Pa6ent   name  

phone#  

Exam  Room  

room#  

5
 Initials:

(b) (12 points) Convert the E/R diagram below to a database schema. Indicate the keys for each
table in your answer. You do not need to write SQL DDL commands.

a   b   c   d  

A   R   B  

isa   f   g   h  

e   C   S   D  

Solution:
A ( c , a , b )
B ( c , d )
C ( a , c , e )
D ( g , h )
S ( a , c , g , f )

Explanation:
Entity A is weak and dependent on B. Hence, the key of A needs to include the key of B.
C is a refinement of A and, hence, needs to include the key of A.

6
 Initials:

(c) (6 points) Suppose we are told that R(A, B, C, D) is in BCNF, and that 3 out of the 4 FDs
listed below hold for R. Choose the FD that R does not satisfy, and explain your reasoning.

1 : A → BCD
2 : BC → A
3 : CD → B
4: D→C

Solution:
The problem states that only 3 out of the given 4 FDs hold for R. So, there are 4 possibilities:
• If 2, 3, 4 are the ones that hold, then D is the only key, and therefore 2 would violate BCNF,
so it’s not a possibility.
• If 1, 3, 4 are the ones that hold, then the only key is A. But then both 3 and 4 would be in
violation of BCNF, so this is not a good choice either.
• If 1, 2, 4 are the ones that hold, then the keys are A, and BC. But then both 4 would violate
BCNF.
• If 1, 2, 3 are the ones that hold, then A, BC and CD are all keys, so none of them violates
BCNF. Therefore, this is the right choice. So, 4 is the FD that doesn’t hold based on the
problem description.
(Note that all 4 FDs could hold without violation of BCNF, but the problem definition tells you
that only 3 hold, and under this assumption, there is only one possibility as listed above)

7
 Initials:

3. (18 points) Logging and Recovery

Your database server crashed due to a power outage. After rebooting, you find the following log and
checkpoint information on disk, and begin the recovery process. We assume that STEAL/NO FORCE
policy is used, and we use the ARIES method for recovery.

LSN Record prevLSN undoNextLSN

30 update: T3 writes P5 null -
40 update: T4 writes P1 null -
50 update: T4 writes P5 40 -
60 update: T2 writes P5 null -
70 update: T1 writes P2 null -
80 Begin Checkpoint - -
90 update: T1 writes P3 70 -
100 End Checkpoint - -
110 update: T2 writes P3 60 -
120 T2 commit 110 -
130 update: T4 writes P1 50 -
140 T2 end 120 -
150 T4 abort 130 -
160 update: T5 writes P2 Null -
180 CLR: undo T4 LSN 130 150 50

Transaction Table at time of checkpoint Dirty Page Table at checkpoint

Transaction ID lastLSN Status Page ID recLSN
T1 70 Running P5 50
T2 60 Running P1 40
T3 30 Running
T4 50 Running

(a) (2 points) The log record at LSN 60 denotes an update to page P5 by transaction T2. Was this
update to page P5 successfully written to disk? If yes, at what point could that have happened?

Solution:
The update at LSN 60 may have been written to disk; the log entry was flushed before the write
itself. It was not yet flushed at the time of the checkpoint, but may have been flushed later.

(b) (2 points) The log record at LSN 70 denotes an update to page P2 by transaction T1. Was this
update successfully written to disk? If yes, at what point? Briefly explain your answers.

Solution:
The update at LSN 70 was flushed to disk before the Begin Checkpoint (LSN 80). We know this
because it is not in the dirty page table at the time of the checkpoint.

8
 Initials:

(c) (6 points) At the end of the Analysis phase, what will the Transaction and Dirty Page Tables
look like? Populate your answers in the tables below.

Solution:

Transaction ID lastLSN Status Page ID recLSN

T1 90 Running P1 40

T3 30 Running P2 160

T4 180 Aborting P3 90

T5 160 Running P5 50

(d) (6 points) At which LSN in the log will REDO begin? Which log records will be redone? List
their LSNs, and briefly justify your answers.

Solution:

Redo should begin at LSN 40, the smallest of the recLSNs in the dirty page table. The following
log records should be redone:
40, 50, 60, 90, 110, 130, 160, 180
30 is skipped because it precedes LSN 40. 70 is skipped because P 2.recLSN = 160 > 70. Entries
that are not updates are skipped. The CLR record is not skipped, nor is the LSN that it undoes.

9
 Initials:

(e) (2 points) For each of the following questions circle the right answers.
i. In an UNDO only logging scheme, what buffer management policies apply? (circle one of
STEAL or NO STEAL, and one of FORCE or NO FORCE)

STEAL / NO STEAL FORCE / NO FORCE

ii. During recovery with REDO only logging, we cannot use a checkpoint for which we see a
<START CKPT(. . . Ti . . .)> record but no corresponding <END CKPT> record.

TRUE / FALSE

Solution:
i. STEAL and FORCE, ii. TRUE

10
 Initials:

4. (22 points) Concurrency Control

In the schedules given below, the label Ri (X) indicates a read of element X by transaction Ti , and
Wi (X) indicates a write of element X by transaction Ti .

(a) (4 points) Draw the precedence graph for schedule 1. Is schedule 1 conflict-serializable? If so,
what order of the three transactions defines a conflict-equivalent serial schedule?

Schedule 1
R2 (A) R1 (C) R2 (B) W2 (B) R3 (B) R1 (A) R3 (C) W3 (C) W1 (A)

Solution:
Schedule 1 is conflict-serializable because the precedence graph has no cycles. There is an arrow
T2 → T1 because of the conflict R2 (A) . . . W1 (A). There is an arrow T2 → T3 because of the
conflict W2 (B) . . . R3 (B). There is an arrow T1 → T3 because of the conflict R1 (C) . . . W3 (C).
The only possible conflict-equivalent serial schedule is (T2 , T1 , T3 )

T1

T2 T3

(b) (4 points) Draw the precedence graph for schedule 2. Is schedule 2 conflict-serializable? If so,
what order of the three transactions defines a conflict-equivalent serial schedule?

Schedule 2
R2 (A) R1 (C) R2 (B) R3 (B) W2 (B) R1 (A) R3 (C) W3 (C) W1 (A)

Solution:
Schedule 2 is not conflict-serializable because the precedence graph has a cycle. There is an arrow
T2 → T1 because of the conflict R2 (A) . . . W1 (A). There is an arrow T3 → T2 because of the
conflict R3 (B) . . . W2 (B). There is an arrow T1 → T3 because of the conflict R1 (C) . . . W3 (C).

T1

T2 T3

11
 Initials:

(c) (9 points) Can the two schedules from (a) and (b) occur under (non-strict) 2 Phase Locking? If
yes, then add the proper lock/unlock actions to the corresponding schedule(s) in a way compliant
with (non-strict) 2PL. Use Li (X) to denote transaction i locking element X, and Ui (X) to denote
transaction i releasing the lock on X. Assume only exclusive locks are used, and make sure that
no locks remain held at the end of the schedule.
The schedules below are the same ones as in the previous page, and repeated here for convenience
with dotted spaces in-between the actions. You can use this dotted space to add the proper lock
and unlock actions, or if you prefer you can rewrite the schedules including the locking actions in
the empty space further down the page.
If you believe a schedule cannot occur with 2PL, give a brief explanation.

Solution:

There are many correct answers to this question. As long as each piece of data is locked before
used, each lock is not simultaneously held by multiple transactions, and no transaction locks data
after it begins unlocking, the answer is correct.

Schedule 1

L2 (A) R2 (A) L1 (C) R1 (C) L2 (B) R2 (B) W2 (B) U2 (B) U2 (A) L3 (B) R3 (B)

L1 (A) R1 (A) U1 (C) L3 (C) R3 (C) W3 (C) U3 (C) W1 (A) U1 (A) U3 (B)

Schedule 2 cannot be produced by 2PL, as it is not conflict-serializable.

12
 Initials:

(d) (5 points) Circle TRUE or FALSE to reflect the validity of the following statements.
i. For any schedules S1 and S2 , if S1 and S2 are conflict serializable, then S1 and S2 are conflict
equivalent.1

TRUE / FALSE

ii. A SIX lock is compatible with IS and IX locks, i.e. if a transaction holds a SIX lock on an
object, then another transaction can take an IS or IX lock on the same object.

TRUE / FALSE

iii. An IX lock is compatible with an IS lock, i.e. if a transaction holds an IS lock on an object,
then another transaction can take an IX lock on the same object.

TRUE / FALSE

iv. Strict 2PL prevents deadlocks.

TRUE / FALSE

v. In timestamp-based concurrency control, if a transaction gets aborted, it will be restarted

with a new timestamp.

TRUE / FALSE

Solution:

i. FALSE, ii. FALSE, iii. TRUE, iv. FALSE, v. TRUE

1 Two schedules are conflict equivalent, if every pair of conflicting actions appears in the same order in both schedules.

13
Initials:

EXTRA PAGE

14
 CSE 444 Midterm Exam

November 13, 2009

Name ________Sample Solution__________

Question 1 / 24

Question 2 / 22

Question 3 / 22

Question 4 / 12

Question 5 / 20

Total / 100

CSE 444 Midterm, Nov. 13, 2009 Sample Solution Page 1 of 8

Question 1. SQL (24 points, 8 each part) With the holiday gift‐giving season upon us, we’d like to set
up a database to keep track of different products and stores that stock them. Our database has the
following tables:

PRODUCT(pid, pname, manufacturer, color)

STORE(sid, sname, address, city, phone)
INVENTORY(pid, sid, price, quantity)

Table PRODUCT has a unique product id number, pid, for each product, and gives the name,
manufacturer, and color of that product. Table STORE has a unique id number for each store, sid, and
gives the store name, address, city, and phone number. Note that there may be more than one store
with the same name in the same city – for example, there may be several ‘Costco’ stores in ‘Seattle’ –
but the stores will have different addresses and store id numbers. The INVENTORY table gives a list of
which stores carry which products, how many of each product are currently in stock at each store, and
the product price at each different store. The pid and sid entries in this table are foreign keys referring
to the PRODUCT and STORE tables respectively. The quantity in an INVENTORY table entry might be 0 if
a particular store carries a product but currently has none in stock. Also, the same product may have
different prices at different stores.

(a) Write a SQL query that gives a list of store names, addresses, and cities of all stores that sell iPods
(pname is ‘iPod’) that have the color ‘red’. The list should include all such stores, even if they currently
have none in stock. The list should be sorted by city; beyond that, the other information about stores in
a given city may appear in any order.

SELECT s.sname, s.address, s.city

FROM PRODUCT p, STORE s, INVENTORY i
WHERE p.pid = i.pid AND s.sid = i.sid AND p.pname = ‘iPod’ AND p.color = ‘red’
ORDER BY s.city

(continued next page)

CSE 444 Midterm, Nov. 13, 2009 Sample Solution Page 2 of 8

Question 1. (cont) Schemas repeated for convenience:

PRODUCT(pid, pname, manufacturer, color)

STORE(sid, sname, address, city, phone)
INVENTORY(pid, sid, price, quantity)

(b) Write a SQL query that gives the minimum available price of the product named ‘Droid’ from any
store that has one or more ‘Droid’s in stock – don’t consider any store that doesn’t have at least one
‘Droid’ in its current inventory. You only need to produce the minimum price; you don’t need to identify
the store.

SELECT min(i.price)
FROM INVENTORY i, PRODUCT p
WHERE p.pname = ‘Droid’ AND i.pid = p.pid AND i.quantity > 0

(c) Write a SQL query that produces a table showing the availability by city of products with the name
‘Xbox’. The results should have one entry for each city where an ‘Xbox’ is available. The entry for each
city should show the city name, the total number of ‘Xbox’s currently available in all stores in that city
and the average selling price of a ‘Xbox’ in that city (just compute the average of the individual store
prices – don’t consider the number of units in stock at each store). The results should be sorted by city.

SELECT s.city, SUM(s.quantity), AVG(i.price)

FROM PRODUCT p, STORE s, INVENTORY i
WHERE p.pid = i.pid AND i.sid = s.sid AND p.pname = ‘Xbox’
GROUP BY s.city
ORDER BY s.city

CSE 444 Midterm, Nov. 13, 2009 Sample Solution Page 3 of 8

Question 2. Database design (22 points) Continuing with our holiday theme, Santa has been having
trouble keeping track of all of his workshops and elves. He has decided that he needs a database to
organize all this information and has asked you to design it.

(a) Give an E/R diagram that captures the following information.

• Santa has several workshops. Each workshop has a unique name (North Pole, Northwest
Branch, Toy Factory, Hoboken …), and a head elf (the elf in charge of the workshop). A single elf
may be the head of more than one workshop. Although all workshops normally have a head elf,
these positions are vacant from time to time.

• Santa has many elves working for him. An elf only works at one particular workshop, but some
elves are between jobs and are not assigned to any particular workshop at the moment. Each
elf has a name and a unique employee number.

• Some elves supervise other elves, and those elves might themselves have a supervisor who is a
different elf. Most elves are just workers who don’t supervise anyone. An individual elf may
have more than one supervisor if that is appropriate for the particular elf’s job. There are a few
elves that don’t have a supervisor. An elf who is the head of a workshop (see the first paragraph
above) does not necessarily supervise other elves, but he/she might do so.

Draw your E/R diagram below:

name
id name

head

Workshop
Elf
works‐at

supervises

(continued next page)

CSE 444 Midterm, Nov. 13, 2009 Sample Solution Page 4 of 8

Question 2. (cont.) (b) Give a relational schema that captures your E/R diagram from part (a). You
should give the table and attribute names and clearly indicate which attributes are keys and foreign keys
in the various tables. Your tables should be in BCNF (i.e., contain no bad anomalies or functional
dependencies), but you do not need to demonstrate this or prove that your design is in BCNF. You do
not need to give SQL CREATE TABLE statements for your tables.

Notation: tbl.key means a foreign key in the given table.

Elf(id, name, Workshop.name)

Workshop(name, Elf.id)

Supervises(Elf.id, Elf.id)

CSE 444 Midterm, Nov. 13, 2009 Sample Solution Page 5 of 8

Question 3. BCNF (22 points) Consider the relation R(A, B, C, D, E) that has the following functional
dependencies:

AB → C
CD → E
DE → B

Decompose this relation, if necessary, into a collection of relations that are in BCNF. For full credit,
show your work, the intermediate decomposition steps, and show which dependency violations you are
correcting at each step.

There are three different ways to get a BCNF decomposition, depending on which dependency is used
first.

I. Use AB+ = {A, B, C} to decompose R into R1(A, B, C) and R2(A, B, D, E). Table R2 is not in BCNF
because DE+ = {B, D, E} ≠ {A, B, D, E}. Use this to decompose R2 into R21(B, D, E) and R22(A, D, E)

II. Use CD+ = {B, C, D, E} to decompose R into R1(B, C, D, E) and R2(A, C, D). R1 is not in BCNF because
DE+ = {B, D, E} ≠ {B, C, D, E}, so decompose it to get R11(B, D, E) and R12(C, D, E).

III. Use DE+ = {B, D, E} to decompose R into R1(B, D, E) and R2(A, C, D, E). R2 is not in BCNF beause of
CD+ = {C, D, E}, so decompose it to get R21(C, D, E) and R22(A, C, D).

CSE 444 Midterm, Nov. 13, 2009 Sample Solution Page 6 of 8

Question 4. Concurrency Control / Serialization (12 points) We have looked at several ways a database
scheduler can manage transactions to either ensure that their operations are consistent with some
conflict‐serializable schedule, or detect operations that would violate serialization and either delay or
roll back one or more transactions to prevent the problem.

Each of the different concurrency control schemes determines the effective serial order of transactions
based on some operation(s) issued by the transactions. For example, one possibility is to ensure that
the effective serial order is the same order in which the transactions issue their first read operation. Or
the serial order might reflect the order in which the transactions commit or abort.

The significant events that occur during a transaction include some or all of the following:

(a) Start transaction

(b) First read operation issued by the transaction
(c) Last read operation issued by the transaction
(d) First write operation issued by the transaction
(e) Last write operation issued by the transaction
(f) All write operations have completed (i.e., are written to stable storage)
(g) First lock acquired by transaction
(h) Last lock acquired by transaction
(i) First lock released by transaction
(j) Last lock released by transaction
(k) Beginning of validation phase
(l) End of validation phase
(m) Transaction commit or abort (rollback) issued

Write the correct letter in front of each concurrency control scheme below to indicate the event that
determines the effective serial order of transactions using that scheme. If a scheme doesn’t absolutely
guarantee to produce a serializable schedule, use the answer that best determines the effective order of
the transactions.

_____i_____ Two Phase Locking (2PL)

_____a_____ Timestamps

_____l_____ Validation

_____a_____ Snapshot Isolation

CSE 444 Midterm, Nov. 13, 2009 Sample Solution Page 7 of 8

Question 5. ARIES (20 points) After a crash, an ARIES log contains the following entries. There are no
checkpoints in the log, so this is all of the information that is available.

LSN prevLSN Transaction ID Type pageID Data

1 0 T1 U P2 Y
2 0 T2 U P2 Y
3 2 T2 U P1 X
4 3 T2 C ‐‐ ‐‐
5 0 T3 U P2 Y

(Note: the log entries in this question go from top to bottom, unlike the example in class where they ran
from left to right.)

(a) Using the information in this log, show the results of the analysis pass that is the first part of the
ARIES crash recovery. Fill in the transaction table and dirty page table below with the information
discovered during the analysis pass. If an entry in a table is created and then later changed or deleted
during the analysis pass, show the entry and draw a line through all or part of it as appropriate to
indicate the changes or deletions.

Transaction table Dirty page table

Transaction lastLSN Page RecoveryLSN

T1 1 P2 1

T2 2, 3, 4 P1 3

T3 5

(b) What is the FirstLSN, i.e., the LSN of the first log entry that needs to be redone during the redo
phase of the crash recovery that follows the analysis?

(c) Which transaction or transactions are “loser” transactions that need to be undone during the undo
phase of the crash recovery that follows the redo phase?

T1, T3

CSE 444 Midterm, Nov. 13, 2009 Sample Solution Page 8 of 8

Name:

CSE 444, Fall 2010, Midterm Examination

10 November 2010

Rules:

• Open books and open notes.

• No laptops or other mobile devices.
• Please write clearly.
• Relax! You are here to learn.

Question Max Grade

1 30
2 30
3 40
Total 100

1
Initials:

1. (30 points) SQL

Consider a database with the following three relations.

Sensor( sid, type, location, description )

Temperature ( sid, time, tempvalue )
Pressure ( sid, time, pressvalue )

Temperature.sid is a foreign key that references Sensor.sid.

Pressure.sid is a foreign key that references Sensor.sid.

This database holds a set of sensor readings (temperature and pressure) recorded by a set of sensors
deployed at various locations. Some sensors record only temperature readings. Others record only
pressure readings. Some record both. Each sensor produces a reading every minute (pressure reading,
temperature reading, or both). Each sensor is associated with a given type, location, and description.

(a) (8 points) Write a SQL query that returns the type and description of all sensors that have
produced both temperature readings and pressure readings. Each unique combination of type
and description should appear only once in the output.

Solution:

SELECT DISTINCT S.type, S.description

FROM Temperature T, Pressure P, Sensor S
WHERE T.sid = S.sid AND P.sid = S.sid

2
 Initials:

(b) (8 points) Write a SQL query that computes the highest ever pressure reading among all sensors
that recorded both pressure and temperature readings.

Solution:

SELECT MAX(PI.pressvalue)
FROM Pressure PI, Sensor SI, Temperature TI
WHERE PI.sid = SI.sid AND TI.sid = SI.sid

3
 Initials:

(c) (14 points) Consider the set of sensors that record BOTH temperature and pressure (your query
will have to compute that set). Write a SQL query that computes, for these sensors only, the sid
and average tempvalue for all sensors that recorded the highest ever pressure reading. Note that
more than one sensor could have recorded the same, highest pressure reading.

Solution:
We can use our query from the previous question in a subquery to find the sid and average
tempvalue for all sensors that produced such a high pressure reading:
SELECT S.sid, AVG(T.tempvalue)
FROM Pressure P, Sensor S, Temperature T
WHERE P.sid = S.sid AND S.sid = T.sid
AND P.pressvalue >= (SELECT MAX(PI.pressvalue)
FROM Pressure PI, Sensor SI, Temperature TI
WHERE PI.sid = SI.sid AND TI.sid = SI.sid)
GROUP BY S.sid

Alternate solution:
SELECT S.sid, AVG(T.tempvalue)
FROM Pressure P, Sensor S, Temperature T
WHERE P.sid = S.sid AND S.sid = T.sid
AND P.pressvalue >= ALL (SELECT PI.pressvalue
FROM Pressure PI, Sensor SI, Temperature TI
WHERE PI.sid = SI.sid AND TI.sid = SI.sid)
GROUP BY S.sid

4
 Initials:

2. (30 points) Conceptual Design

(a) (12 points) Draw an E/R diagram describing the following domain:
• Users have a single attribute uid (key).
• AnonymousUsers are a type of Users with attribute country.
• NamedUsers are a type of Users with attributes name and email.
• Blogs have attributes bid(key), title, and topic.
• A NamedUser can be a friend-of zero or more other NamedUsers. Each friend-of rela-
tionship has an associated start date.
• A NamedUser can own many Blogs, but each blog is owned by exactly one user.
• A User can post to zero or more Blogs. Each post has a given content. Many users can
post to a blog.
Your answer should consist of an E/R diagram, which includes entity sets, attributes, relationships,
ISA relations. Indicate the type of each relationship with appropriate arrows (one-one, one-many,
etc.).

Solution:

uid

date
content Users

Friend
isa isa ‐of

post Anonymous Named

country
Users Users name

own
email

Blogs

bid 6tle topic

5
 Initials:

(b) (8 points) Consider the following relational schema and set of functional dependencies. (a)
List all superkey(s) for this relation. (b) Which of these superkeys form a key (i.e., a minimal
superkey) for this relation? Justify your answer in terms of functional dependencies and closures.

R(A,B,C,D,E) with functional dependencies CD → E and A → B.

Solution:
(a) A superkey is a set of attributes X s.t. X + = all attributes.

From the FDs above, we can derive:

{A, B, C, D, E}+ = {A, B, C, D}+ = {A, C, D, E}+ = {A, C, D}+ = {A, B, C, D, E}

Hence,
{A, B, C, D, E}, {A, B, C, D}, {A, C, D, E}, and {A, C, D} are all superkeys.

(b) A key is a set of attributes which form a superkey and for which no subset is a superkey. In
our example, {A, C, D} is the only key.

6
 Initials:

(c) (10 points) Decompose R into BCNF. Show your work for partial credit. Your answer should
consist of a list of table names and attributes and an indication of the keys in each table (underlined
attributes).

Solution:
Both functional dependencies violate BCNF.
Try {A}+ = {A, B}. Decompose into R1(A,B) and R2(A,C,D,E).

R1 has two attributes, so it is necessarily in BCNF.

R2 is not in BCNF, since {C, D} is not a key and we have CD → E.

Try {C, D}+ = {C, D, E}. Decompose into R3(C, D, E) and R4 (A, C, D )

End result: R1(A,B), R3(C, D, E), and R4 (A, C, D )

7
 Initials:

3. (40 points) Transactions

(a) (6 points) In the ARIES method, assuming NO checkpoints have been used, explain what
happens during the ANALYSIS pass of recovery. Your answer should indicate at least (1) what
part of the log the system reads and in what direction and (2) what data structures the system
rebuilds.

Solution:
In the absence of checkpoints, the analysis pass reads the log from the beginning to the end. It
rebuilds the Dirty Pages Table and the Transactions Table to determine the state of the system
as of the time of the crash. It rebuilds these data structures by updating them according to the
log records that it encounters during the forward scan.

(b) (6 points) After the analysis pass, the protocol performs a REDO pass. Explain (1) where does
REDO start reading the log and in what direction it reads the log, (2) what happens during the
REDO pass.

Solution:
The REDO pass begins at the log record whose LSN corresponds to the earliest recoveryLSN of
all the entries in the Dirty Page Table. From that point, the REDO pass redoes the updates of
all transactions (committed or otherwise). At the end of this pass, the database is in the same
state as it was right before the crash.

8
 Initials:

(c) (6 points) The last pass during recovery is the UNDO pass. Explain (1) where does the UNDO
start reading the log and in what direction it reads the log, (2) what happens during the UNDO
pass.

Solution:
The UNDO pass scans backward from the end of the log. The pass undoes the updates by all
transactions that had not committed by the time of the crash. These transactions can be found in
the Transactions Table rebuilt during the analysis pass. All updates are undone unconditionally
(since the REDO pass ensured that all logged updates have been applied to affected pages).
For each update that is undone, the undo operation is logged with a Compensation Log Record
(CLR).

9
 Initials:

(d) (7 points) Consider the following two transactions and schedule (time goes from top to bottom).
Is this schedule conflict-serializable? Explain why or why not.

Transaction T0 Transaction T1
r0 [A]
w0 [A]
r1 [A]
r1 [B]
c1
r0 [B]
w0 [B]
c0

Solution:
The schedule is not conflict serializable because the precedence graph contains a cycle. The graph
has an edge T0 → T1 because the schedule contains w0 [A] → r1 [A]. The graph has an edge
T1 → T0 because the schedule contains r1 [B] → wo [B].

Because of W0[A] R1[A]

T0 T1

Because of : R1[B] W0[B]

10
 Initials:

(e) (7 points) Show how 2PL can ensure a conflict-serializable schedule for the same transactions
above. Use the notation Li [A] to indicate that transaction i acquires the lock on element A and
Ui [A] to indicate that transaction i releases its lock on A.

Solution:

Multiple solutions are possible.

Transaction T0 Transaction T1
L0 [A]
r0 [A]
w0 [A]
L1 [A] → blocks
L0 [B]
r0 [B]
w0 [B]
U0 [A]
U0 [B]
c0
L1 [A] → granted
r1 [A]
L1 [B]
r1 [B]
U1 [A]
U1 [B]
c1

11
 Initials:

(f) (8 points) Show how the use of locks without 2PL can lead to a schedule that is NOT conflict
serializable.

Solution:

Transaction T0 Transaction T1
L0 [A]
r0 [A]
w0 [A]
U0 [A]
L1 [A]
r1 [A]
U1 [A]
L1 [B]
r1 [B]
U1 [B]
c1
L0 [B]
r0 [B]
w0 [B]
U0 [B]
c0

12