Q157: About Spinlocks

Spinlock is a lock which causes a thread trying to acquire it to simply wait in a loop while repeatedly checking if the lock is available. This is similar to busy-waiting, since the thread is active but is not doing any useful work.

Question #157: Which of the following makes sense regarding spinlocks?

Options:

1. Spinlock is more useful for uni-processor systems, than multiprocessor systems.

2. Spinlock is more useful for multiprocessor systems, than uni-processor systems.

3. Spinlocks can be useful as they avoid the overhead of context switch and CPU scheduling.

4. Acquiring a lock cannot happen without spinlocks.

Solution: In a uni-processor system, if a thread has nothing better to do than just wait, it makes more sense to release the control of processor and go to sleep, so that some other thread/process can take over the processor and do some useful work. However, in a multiprocessor system, it’s okay for a thread to waste a few CPU cycles in spinlock, since other processes can still run on the other processors during that time. Imagine a scenario where process A is running and it  holds lock L, but a higher priority process B comes along, causes context switch and gets hold of the CPU. Now, process B starts running but needs lock L. If process B goes into spin lock, then all the CPU cycles are wasted as unless process A is allowed to run, it cannot release L. Hence option 2 is right over option 1.

Option 3 is also correct, since if a process has to wait for a short time, and then run again, it makes sense to use spinlocks than to do context switches. Option 4 is false, since it is not necessary to implement spinlock for acquiring locks. A thread can go to sleep if a lock is unavailable and be woken up when a lock becomes available again.

So both options  2 & 3 are correct here.

Q152: CPU Idle time in SRTF scheduling

Question 152: Consider 3 processes, all arriving at time zero with total execution time of 10, 20, and 30 units respective each process spends the first 20% of execution time doing I/O, the next 70% of time doing computation and last 10% of time doing I/O again. OS uses SRTF scheduling algorithm. Assuming I/O of processes can be overlapped, what percentage of CPU time will remain idle?

Options:

1. 0%
2. 30%
3. 6%
4. 4%

Solution: The correct answer is the third one. Processes will run as follows:

0-2: Idle

2-9: P1

9-23: P2

23-44: P3

44-47: Idle

Idle time = 5 / 47 * 100 = 10.6%

Q104: Performance of Round Robin Scheduling

Question #104: The performance of Round Robin Scheduling depends on:

Options:

1. Size of the processes
2. CPU bursts of the processes
3. I/O bursts of the processes
4. The size of the time quantum

Solution: The correct answer is last one, because if quantum is big then RR behaves like FCFS, and if it is too small then context switching overhead increases.