Skip to content
This repository

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP

CPU Scheduling Simulation for Operating Systems

branch: master
readme.md

Scheduling Project

Simulate three CPU scheduling techniques:

  1. Multilevel Feedback Queue (MLFQ) [1]; MLFQ utilizes multiple queues and allows processes to move between queues dynamically. In this approach, processes "find their own level" based on their CPU burst . Processes that have a larger CPU burst moves to a lower-priority queue.

  2. Lottery Scheduling [2], each process is given some lottery tickets. A lottery is held at regular intervals and the winner is determined by selecting a ticket at random. The winning process gets to be executed next. A process will have a chance of winning proportionate to the number of tickets it has. To increase the chances of winning, higher priority processes can get more tickets. Lottery Scheduling guarantees a non-zero probability for any process to get executed and hence solves the starvation problem.

  3. Fair-Share Scheduling [3]: divide CPU time evenly among users and then among processes. For example, if we have two users X and Y in the system and each has one process, A and B respectively. Then these will be scheduled in this fashion: A B A B A B …., resulting in each user getting 50% of the CPU time. However, if user X has two processes A and C, with 50% to each user, the processes will be scheduled as A B C B A B C B…., then A will take 25%, B 50% and C 25%. If on the other hand a third user Z has a process C then each user will take 1/3 of CPU time and the processes are scheduled as A B C A B C….., with A taking 33.3%, B 33.3% and C 33.3%.

Test and compare the Performance of the three schedulers using a performance metric(s) (e.g. waiting time) of your choice under different CPU-I/O bound loads

How would the three schedulers perform on multicore processors?

Submit a report that explains your • implementation details • assumptions • values for parameters: e.g. quantum value, number of tickets,...etc., • explanation of testing method, creation of jobs: random, poisson arrival -exponential departure,...etc. CPU-bound or I/O bound
• discussion of the results

Submit your code

References:

[1] Hoganson K., Reducing MLFQ Scheduling Starvation with Feedback and Exponential Averaging, The Journal of Computing Sciences in Colleges, 25, (2), 196- 202, 2009.

[2] Waldspurger and Weihl, Lottery Scheduling: Flexible Proportional-Share Resource Management, Proceedings of First Symposium on Operating System Design and Implementation, 1-12, 1994.

[3] Henry G., The Fair Share Scheduler, AT&T Bell Lab Technical Journal, 63, (8), 1845-1857 1984.

Something went wrong with that request. Please try again.