Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit

INFO20003 Database Systems

SAMPLE EXAM

Section 1 – ER Modelling

The Australian Department of Health and Human Services (DHHS) has a requirement to store the vaccination record of every individual (known as a patient) that resides in Australia who is eligible for a Medicare card. For families, children under the age of 15 are stored on the family card (refer Figure 1). Each Medicare card number is associated with only one family. Every Medicare card has a 'valid to' date stored as a month and year on the card (refer Figure 1). Each family member holds a position number on the card. For example to identify Jessica Smith (refer  Figure  1) both her Medicare number (1234567890) and position (4) on the Medicare card would be required.

 

Figure 1: Australian Medicare card (DHHS, Australia)

Each Medicare card is attached to one residential address, contact email, and phone number. For all patients listed on a Medicare card we record their gender, birthdate, first name, last name, and if they have any known allergies (e.g. Penicillin, Cortisone, Codeine). If patients do have an allergy we need to know what the allergy is and the reaction (severe, moderate or mild).

The DHHS needs to record mandatory vaccinations (e.g. Measles, Polio, Whooping Cough, Diphtheria, Tuberculosis and Tetanus) as well as optional vaccination types (e.g. HPV, Flu, Hepatitis A, Hepatitis B, Cholera, Typhoid, Yellow Fever).

For each vaccination event that is given to patients we must record the vaccination type, date of vaccination and the vaccination batch number. Vaccine producers can produce many different types of vaccines and each vaccine can have many batches.

Patients can receive their vaccination from any registered doctor. Every doctor is identified by a unique medical practitioner number (MPN). We record the Medical Practitioner's title (Dr, Mr, Mrs, Ms, Prof.), first name, last name, registered business address, email, and business phone numbers.

Q1. Draw a conceptual model in either Crow’s Foot or Chen’s notation for this case study (in your script book). Be sure to write down any assumptions you make.

(20 marks)

Section 2 – SQL-DDL

 

Figure 2: Data model for SQL DDL

Q2. Write SQL statements to create the tables for the above data model. Be sure to specify primary and foreign keys. You do not need to specify whether the fields are NULL/NOT NULL. Choose appropriate data types for attributes that are not obvious.

 

(5 Marks)

Section 3 – SQL-DML

Figure  3  shows  the  schema  for  a  small  business  database  that  contains  data  about employees, products, customers and orders.

 

Figure 3: Data model for SQL DML

Write a single SQL statement to correctly answer each of the following questions (3A – 3D). DO NOT USE VIEWS or VARIABLES to answer questions. Query nesting is allowed.

The relations are repeated here for your convenience.

FK

employees (empid, lastname, firstname, hiredate, address, phone, managerid)

FK            FK

orders (orderid, custid , empid , orderdate, shippeddate, freight, shipname) customers (custid, companyname, contactname, address, phone)

FK                 FK

orderdetails (orderid , productid , quantity, discount)

products (productid, productname, unitprice, discontinued)

Q3A. Write a query that returns customers (company names) and the details of their orders (orderid and orderdate), including customers who placed no orders.

(3 marks)

Q3B. Write a query that returns the first name and last name of employees whose manager was hired prior to 01/01/2002.

(4 marks)

Q3C. Write a query that returns customers whose company name is Google’ , and for each customer return the total number of orders and total quantities for all products that were not discontinued (‘1’ means discontinued, ‘0’ not discontinued).

(5 marks)

Q3D. Write a query that returns the ID and company name of customers who placed orders in 2007 but not in 2008.

(8 marks)

Section 4 – Query Processing Joins

Given the schema of Question 3 (SQL-DML), consider the relations Orders and Employees. Imagine that relation Employees has 1,000 pages and relation Orders 50,000 pages. Consider the following SQL statement:

SELECT *

FROM Employees NATURAL JOIN Orders

WHERE freight > 1000;

There are 502 buffer pages available in memory. Both relations are stored as simple heap files. Neither relation has any indexes built on it.

Q4A. What is the cost (in disk I/O’s) of performing this join using the Block-oriented Nested Loops Join algorithm? Provide the formulae you use to calculate your cost estimate.

(3 marks)

Q4B. What is the cost (in disk I/O’s) of performing this join using the Hash Join algorithm? Provide the formulae you use to calculate your cost estimate.

(3 marks)

Q4C. In comparing the cost of different algorithms, we count I/O (page accesses) and ignore all other costs. What is the reason behind this approach?

(2 marks)

Q4D. Which approach should be the least expensive for the given buffer size of 502 pages:

1.         Simple Nested Loops Join

2.         Page-oriented Nested Loops Join

3.         Block-oriented Nested Loops Join

4.         Hash Join

Please write the number of the correct response. No need to provide formulae for question

4D.

 

(2 marks)

Section 5 – Query Processing Indexing

Given the schema of Question 3 (SQL-DML), consider the relation OrderDetails. Imagine that the relation OrderDetails consist of 100,000 tuples stored in pages and each page stores 100 tuples. Imagine that the quantity attribute can take any value between 0 and 20 ([0, 20]), and imagine that discount can take any value between 0 and 100 ([0, 100]). Suppose that the following SQL query is executed frequently using this relation:

SELECT *

FROM OrderDetails

WHERE quantity > 15 AND discount > 90;

Your job is to analyse the following query plans and estimate the cost of the best plan utilizing the information given about different indexes in each part.

Q5A. Compute the estimated result size of the query, and the reduction factor of each filter.  (3 marks)

Q5B. Compute the estimated cost of the best plan assuming that a clustered B+ tree index on quantity is (the only index) available. Suppose there are 200 index pages.

(3 marks)

Q5C. Compute the estimated cost of the best plan assuming that an unclustered Hash index on discount is (the only index) available.

(2 marks)

Q5D. If you are given complete freedom to create one index to speed up this query, which index would be the best one to answer this query? Please give complete information about

the index, e.g. is it clustered or unclustered, is it hash or B+ tree, which attributes will it cover. (2 marks)

Section 6 – Query Optimisation

Given the schema of Question 3 (SQL-DML), consider the relations Employees, Orders and OrderDetails. Imagine that relation Employees has 1,000 pages, relation Orders 5,000 pages, and relation OrderDetails 10,000 pages. Each page stores 100 tuples, and neither relation has any indexes built on it.  Consider the following query:

SELECT *

FROM Employees AS E, Orders AS O, OrderDetails AS OD

WHERE E.empid = O.empid AND O.orderid = OD.orderid;

Q6A. Compute the cost of the plan shown below. NLJ is a Page-oriented Nested Loops Join. Assume that empid is the candidate key of Employees, orderid is the candidate key of Orders, and 100 tuples of a resulting join between Employees and Orders can fit on one page.

 

(4 marks)

Q6B. Would the plan presented below be a valid candidate that System R would consider and compute cost for during query optimisation? Why?

 

(3 marks)

Q6C.  Consider the  query  presented  below.  Does the following  equivalence  class  hold? Yes/No and Why?

SELECT firstname, lastname

FROM Employees NATURAL JOIN Orders NATURAL JOIN OrderDetails

WHERE quantity > 5 AND freight < 100

Πfirstname,  lastname(σquantity> 5 ۸ freight <100(Employees Orders OrderDetails))

↔ σquantity> 5 ۸ freight <100(Πfirstname,  lastname(Employees  Orders  OrderDetails))

(3 marks)

Section 7 – Normalisation

Q7. The table shown  below  is  part of an office  inventory database.  Identify the design problems and write a revised table structure in 3rd  Normal Form (3NF) that corrects those problems. For each step explicitly identify which normal form is violated and briefly explain why.

Write the normalised tables in a textual format, as in:

FK

TableName (PrimaryKey, Column, ForeignKey)

AnotherTable (PrimaryKey, Column, AnotherColumn)

Item ID is the candidate key for this table. Item ID determines Description, Quan, Cost/Unit and Dept, while Dept determines Dept Name and Dept Head.

Item ID

Description

Dept

Dept Name

Dept Head

 

Quan

Cost/Unit

Inventory Value

Jane Thompson

Jane Thompson

Jane Thompson

Ahmad Rashere

(10 marks)

Section 8 – Data Warehousing

You are making a data warehouse for a real estate agency. The company wants to track information about the selling of their properties. Whenever a buyer comes into the office an agent takes that person to a number of properties and deals with the buyer. This warehouse keeps information about the agents (real estate license#, first name, last name, phone #, and branch office), buyers that come in (buyer id, first name, last name, phone #, max price), and property (property#, property address, number of rooms, pool, owner id). The information managers want to be able to find is the number of times a property is viewed, sales price and commission. Sales commission is additional compensation the agent receives for exceeding expectations. The information needs to be accessible by agent, by buyer, by property and for different time (day, week, month, quarter and year).

Q8A. Draw a star schema to support the design of this data warehouse, showing the attributes in each table. Use PK to denote primary key, PFK to denote primary foreign key, and FK to denote foreign  key. You  do  not  need to  specify  data types  nor whether the fields  are

NULL/NOT NULL.

(8 marks)

Q8B. Why are star schemas preferred over relational database designs to support decision making?

(2 marks)


Section 9 – NoSQL and Database Administration

Q9A. There are trade-offs between the principles of ACID and BASE. Discuss the trade-off between availability and consistency for relational and  NoSQL databases.  Illustrate your answer using either the example of Facebook or the example of Twitter.

(5 marks)

Q9B. Vine is a social media sharing service where users can host 6-second video clips within multiple categories (e.g. “Comedy”, “Science”, “Social”). Part of the database schema for the Vine service is given below.

Storage requirements for different data types

Data type

Storage

requirement

(bytes)

INT

4

BLOB

65,535 (Max)

DATETIME

8

There are 15 different categories that users can share videos about and 1 million users to start with.  A  user  posts  5  videos  on  average  per  month. Assume  that  the  average  storage requirement for the BLOB data type is 20,000 bytes.

Estimate the disk space requirements only for the Video table at go-live and after one month of operation.

(5 marks)

Section 10 – Multiple Choice Questions

There are fifteen multiple choice questions. Each question is worth 1 mark. There is only one correct answer per question. Write the question number and your answer in the script book.

Q10A. Which of the following is NOT part of the database development lifecycle?

A)  Implementation

B)  Maintenance

C)  Requirement analysis

D)  First-level support

Q10B. A relation which is NOT in the conceptual and logical models but is made available to users is a:

A)  Data type

B)  View

C)  Revoke

D)  Grant

Q10C. Which of the following is NOT a valid join?

A)  Sort-merge

B)  Hash

C)  Nested loop

D)  Heap

Q10D. Which of the following is NOT a true statement about data and information?

A)  Dates and audio are two examples of data

B)  Information has been processed so that it increases the audience’s knowledge

C)  Information is used to create data

D)  The term datarefers to raw facts

Q10E. Which of the following is NOT true about normal forms?

A)  A table in 1NF must have no multivalued attributes

B)  A table in 2NF must have no multivalued attributes

C)  A table in 2NF must have no transitive functional dependencies

D)  A table in 3NF must have no transitive functional dependencies

Q10F. Which of the following is NOT one of the basic operations of Relational Algebra?

A)  Set-difference

B)  Cross-product

C)  Equality

D)  Union

Q10G. The structure of a Clustered B+ tree Index does NOT contain:

A)  Data pages

B)  Leaf nodes

C)  Internal nodes

D)  Root node

Q10H.  Which of the following is NOT a true statement about projection?

A)  Projection can be accomplished by hashing

B)  The cost of projection by external merge sort depends on the reduction factor

C)  Projection algorithms are designed to remove duplicates from the result set

D)  Projection by external merge sort uses the same sorting approach as sort-merge join

Q10I. Which of the following is NOT a commonly-used level of lock granularity?

A)  Field-level locking

B)  Row-level locking

C)  Table-level locking

D)  Database-level locking

Q10J. Which of the following is NOT a valid type of logical database backup?

A)  Incremental backup

B)  Online backup

C)  Onsite backup

D)  Offline backup

Q10K. Which of the following is NOT a normalization concept?

A)  A.C.I.D.

B)  Partial functional dependency

C)  Candidate key

D)  Determinants

Q10L. Which of the following is NOT a file organisation type?

A)  Hash file organisation

B)  Index file organisation

C)  Heap file organisation

D)  Sorted file organisation

Q10M. Which of the following is NOT an SQL command category?

A)  DDL

B)  DML

C)  DCL

D)  DSL

Q10N. Which of the following is NOT a part of query optimisation?

A)  Plan enumeration

B)  Costing

C)  Building a new index

D)  Result size estimation

Q10O. Which of the following is NOT true about choosing data types?

A)  May help improve query optimisation

B)  Help DBMS store and use information efficiently

C)  Help minimise storage space

D)  May help improve data integrity