Scribd
Upload
74 views

2003 Test and Answers

This document contains 8 questions regarding operating systems concepts such as kernels, scheduling, concurrency, and distributed systems. Question 1 asks about monolithic and microkernels and their properties. Question 2 discusses a C program that uses fork() and produces a process tree. Question 3 covers priorities in real-time scheduling. Question 4 describes how thread suspension can cause deadlocks. Question 5 lists changes needed to support time-sharing. Question 6 is about scheduling algorithms. Question 7 discusses atomic semaphore operations. Question 8 asks about mutual exclusion in distributed systems.

Uploaded by

Belete Belay
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
74 views

2003 Test and Answers

This document contains 8 questions regarding operating systems concepts such as kernels, scheduling, concurrency, and distributed systems. Question 1 asks about monolithic and microkernels and their properties. Question 2 discusses a C program that uses fork() and produces a process tree. Question 3 covers priorities in real-time scheduling. Question 4 describes how thread suspension can cause deadlocks. Question 5 lists changes needed to support time-sharing. Question 6 is about scheduling algorithms. Question 7 discusses atomic semaphore operations. Question 8 asks about mutual exclusion in distributed systems.

Uploaded by

Belete Belay
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

OS test questions and answers

Question 1.
There are two major design types for operating system kernels: monolithic kernels and microkernels.
Which of these types better satisfies the following requirements? If both are equally as good write
both. Justify your answers.
a) Convenient access to operating system data structures.
Monolithic – access from any part of the kernel to any other part is possible. With a microkernel
access is via message passing.

b) Addition of new operating system components.


Microkernel – no (or little) modification is required in the kernel the new service is in user space.

c) Modification of operating system components.


Microkernel – no (or little) modification is required in the kernel. Modifying a user mode service
can be done more easily. It should be possible to modify without a need to restart the system.

d) Security and reliability.


Microkernel – services are isolated from each other. If one service crashes it does not directly
affect others. Another acceptable answer on security would be – both, if implemented well.
8 marks
Question 2.
In the following Linux program written in C the function getpid() returns the process id of the current
process.
The line
printf("%d. childpid: %d\n", i, childpid);
is equivalent to the Java
System.out.println(i+". childpid: "+childpid);

/* testForkExample.c */
main() {
int i = 0;
printf("main pid: %d\n", getpid());
while (i < 3) {
int childpid = fork();
if (childpid != 0)
printf("%d. childpid: %d\n", i, childpid);
i++;
}
}

The program produced the following output when run from a terminal on my machine:

[robert@icicle Documents]$ testForkExample


main pid: 3260
0. childpid: 3261
1. childpid: 3262
2. childpid: 3263
[robert@icicle Documents]$ 1. childpid: 3264
2. childpid: 3265
2. childpid: 3266
2. childpid: 3267
a) Draw a process tree showing each of the processes produced with links between parent and
child processes. (There is more than one possible correct answer.)

pid:3260

pid:3261 pid:3262 pid:3263

pid:3264 pid:3265 pid:3266

pid:3267
Other acceptable answers have 3265, 3266 and 3267 swapped amongst themselves. The others
must be in these places.
6 marks
b) Explain why the last four lines of output are produced after the command line prompt
[robert@icicle Documents]$ reappears.
The prompt can reappear as soon as the testForkExample process itself has finished. The shell
waits for any process it starts to finish it does not wait for children of that process to finish. All
other lines of output come from the children.
2 marks
Question 3.
a) In the lectures it was mentioned that priorities make real-time scheduling more difficult. Give a
reason why priorities are still used in real-time systems

Priorities can be used to indicate how important tasks are. In this way we can ensure that more
important tasks are handled before less important ones.
2 marks
b) Here is a diagram from the textbook:

Briefly describe each of the following terms shown in the diagram:


conflicts Conflicts are conditions that cause the required response to be postponed,
e.g., some required resource is unavailable and a process needs preempting.
dispatch Making the required response change state to running. At the end of this
time the real-time process is running.
dispatch latency The total time from when the correct real-time process is identified by the
interrupt handler through to the start of execution of that process.
response interval The total time from when the event occurs until the event has been
completely handled.
8 marks
Question 4.
The Java thread method suspend() is deprecated because it can produce deadlocks. Describe a
situation where the use of suspend results in a deadlock.
Any situation where the thread that is going to call resume on the suspended thread itself blocks
waiting for a lock currently held by the suspended thread.
5marks
Question 5.
In the early years of digital computers, programmers could work interactively on the machines. For
many years following that time programmers had to submit their programs for processing at a later
time and could no longer control the machines directly. What changes needed to be made to the
hardware and operating systems that allowed programmers safe interactive access to the computers
again on time-sharing systems? Also say why the changes were needed.
Memory protection. So that a programmer working on the computer doesn't corrupt memory being
used by another user or program.
Protected processor mode. To stop the programmer directly accessing resources that need
protection either because they are shared or because they are dangerous if used incorrectly.
Plenty of resources (especially more memory), time-sharing systems cannot maximise resource
usage without penalizing usability.
Terminals for all users. Obvious need for interactive computing.
Scheduling needed to change (or be added). Want to provide good service to the user rather than
maximising the use of the hardware.
Security. Apart from the hardware changes above, interactive computing allows greater
possibilities for security violations, so file system and other security needed adding or improving.
etc.
6 marks
Question 6.
a) A scheduler uses multi-level feedback queues with the following characteristics:
Level Time-slice When dispatched
1 10 msec
2 40 msec Only when level 1 is empty
3 Unlimited Only when levels 1 and 2 are empty

Processes enter the scheduler in the level 1 queue and move down a level if they completely use their
time-slices. Processes do not move up levels. Preemption only occurs when the time-slice is
completed.
This list of processes shows CPU burst times. The processes start in this order in the level 1 queue.

Process name CPU burst time


A 8
B 300
C 60
D 12
E 5
F 20

Using the information above complete this table showing: the order of process scheduling; the queue
the process was selected from; and how long each process runs.

Process Queue Run time


A 1 8
B 1 10
C 1 10
D 1 10
E 1 5
F 1 10
B 2 40
C 2 40
D 2 2
F 2 10
B 3 250
C 3 10

5 marks
b) There is something very wrong with this scheduler. What is it and how can it be fixed?

When processes descend to level 3 the system can be locked up by a level 3 process running in an
infinite loop.
Fix it by giving level 3 processes a long but fixed time-slice and scheduling them round-robin.
2 marks
Question 7.
The semaphore wait and signal operations are defined as indivisible (or atomic) operations.
a) Why do they have to be indivisble?
If they weren't indivisble it would be possible for multiple waits and signals to occur simultaneously
(on a multiprocessor) or at least interleaved with each other. This would corrupt the value of the
semaphore and prevent it working properly.
4 marks
b) Because they are indivisible does that mean that all other processes running on a multiprocessor
system must stop when a wait or signal operation is executed? Explain why or why not.
No. Strictly only processes accessing the same semaphore at that time need to stop. Any processes
not wanting to change or access the semaphore should be allowed to continue.
4 marks
Question 8.
a) Describe a scheme to implement mutual exclusion in a distributed environment. Say how a process
requests exclusive access to the resource, how that exclusive access is maintained and how the
resource is made available to a new process when the current one has finished with it.

Either a centralized or distributed solution is acceptable. See the notes.


6 marks
b) What are two problems with the scheme you described in a)?
If the site hosting the process with the resource goes down any solution will need to account for
that.
In centralized systems the coordinator process going down is a problem.
Network failures will be a problem in any solution.
2 marks

You might also like