2 - Design

Question 1

realiability + metric

Answer 1

The notion of reliability refers to the ability of a system to offer a continuous service (without
failures) under specific stated conditions during a specified period of time.

Metric: Mean Time Between Failures (MTBF)

MTBF = sum(start_down_time - start_uptime)/number_failures

Question 2

Design faults are _________

Answer 2

permanent

Question 3

Simulation of the design of a program can be used to eliminate all requirements
faults.

Answer 3

False, because 1) no method can eliminate all the faults in a program, 2) missing
requirements cannot be detected, and 3) usually it is not possible to explore all the states of a
program by simulation.

Question 4

Voting systems (e.g. triple modular redundancy) can be used to tolerate transient
faults but not permanent faults.

Answer 4

We focus on permanent faults part of the statement.
True, if the fault occurs in software. All the replicas will present the same (design) fault,
therefore redundancy would not help.
False, in the case of hardware, as long as long as the fault is not present in all the replicas.

Question 5

A method for fault forecasting is to build in adaptive load control in the design of
the system.

Answer 5

True, switching load from one server to another (as in the case of the Google mail server)
would need identification of current/future load and adaptation to it.

Question 6

Explain whether production defects in microchips are an example of faults, errors or
failures.

Answer 6

The production defects in a microchip can be a failure and a fault in two different contexts:
 Failure of the production line of the microchips.
 Fault of an incorrectly delivered service that uses these microchips.
Nevertheless they cannot be an error, because they are not the symptom of a problem.

Question 7

Describe the relation between “degraded mode” and “system failure”. You may use an
example to explain whether these terms are synonyms or have differences.

Answer 7

System failure implies that the main functionality of a system is lost, e.g. the engine of
a car breaks and the car cannot transport anyone.
Instead, degraded mode implies that only a subset of the system functionality is
available or that it is available with reduced performance, e.g. a cellular network that
because of congestion only allows to send SMSs, but not to establish a call.
Therefore, system failure is the worst scenario, since the service is not provided at all
after the failure, while in degraded mode the system still provides part of the services.

Question 8

Explain the notion of graceful degradation and give one example of it in a real application
setting.

Answer 8

Graceful degradation is the capability of a system to offer a reduced service or
performance in case of failure of some of its components.
An example can be a communications network that has some of its links down and it
still can offer the service, but with a reduced bandwidth available and higher latency,
e.g. Internet.

Question 9

Redundancy in hardware through triple modular redundancy does not increase the
response time of an application compared to a non-replicated solution.

Answer 9

True, if the time spent for voting over the results is negligible.

Question 10

TCP employs redundancy in data when a message is retransmitted to achieve
reliable communication.

Answer 10

False, TCP includes a checksum to detect errors, but not to correct them.

Question 11

TCP employs redundancy in time when a message is retransmitted to achieve
reliable communication.

Answer 11

True, TCP retransmits segments of data that has been lost or corrupted.