Distributed Data Concepts

Question 1

Transaction

Answer 1

Completes changes successfully or not at all. This lets multiple users access the database without fear of data inconsistencies.

Question 2

Rollback

Answer 2

Undoes all changes in a transaction due to an error

Question 3

True or false: A transaction can only have one read operation and one write operation

Answer 3

False! A transaction can have as many operations as needed.

Question 4

Online Transaction Processing (OLTP)

Answer 4

Real-time access to data

Question 5

Online Analytical Processing (OLAP)

Answer 5

Involves fewer, more intensive transactions (e.g. banks processing large amounts of money) than OLTP

Question 6

Lost update

Answer 6

When multiple users attempt a transaction at the same time, only one completes successfully

Question 7

Dirty read

Answer 7

Data loss (e.g. due to a power cut)

Question 8

True or false: In the event of data loss, data is always reset to its original value

Answer 8

False! Whether the data is reset depends on whether the changes made were committed beforehand.

Question 9

Inconsistent analysis

Answer 9

One user reads data while another user is updating it, resulting in a mix of old and new values

Question 10

Locking

Answer 10

Transactions lock part of the database before updating via a shared lock (read only) or an exclusive lock (read/write)

Question 11

Timeout

Answer 11

A transaction rolls back after a certain amount of time

Question 12

Deadlock detection

Answer 12

The smallest transaction is rolled back

Question 13

Distributed databases

Answer 13

Data is stored in different physical locations

Question 14

True or false: Distributed databases can lead to performance issues

Answer 14

True! More locations = higher risk of performance issues

Question 15

CAP Theorem

Answer 15

Consistency - transactions ensure this by rolling back if an error occurs
Availability - every query request gets a response
Partition tolerance - coping with network failures/delays

Question 16

True or false: A good database needs all CAP aspects

Answer 16

False! According to Brewer’s Theorem, only 2 CAP aspects are needed at one time (one being partition tolerance).

Question 17

Why is partition tolerance necessary?

Answer 17

Occasional network issues is better than data anomalies

Question 18

Logging

Answer 18

Transactions are logged, containing original and new values in case of rollback

Question 19

Checkpoint

Answer 19

Logs which transactions are running, committed, etc.

Question 20

What are some issues that can damage a database?

Answer 20

o Sudden crash or loss of power to servers
o Hardware problems (e.g. corrupt/broken disks)
o Flood/fire/etc. in server room
o Accidents
o Malicious damage

Question 21

True or false: In the event of damage, all incomplete transactions must be undone

Answer 21

True! These transactions will have neither COMMIT nor ROLLBACK in the log.

Question 22

True or false: If the database is damaged, completed transactions must be undone

Answer 22

False! Completed transactions (that weren’t marked as complete at the last checkpoint) are redone in case the DBMS didn’t finish writing the changes to the disk.

Question 23

Where should archives and log files be stored?

Answer 23

At a different location to the database, preferably not on a server

Question 24

Why is indexing important?

Answer 24

It makes query processing more efficient; without it, a DBMS would have to search a whole table to look for data.

Question 25

Clustered index

Answer 25

Controls how data is stored on a disk

Question 26

Non-clustered index

Answer 26

Contains a pointer (row locator) to each row