COMP SCI 3004/7064 – Operating Systems Assignment 1
Important Notes
• Handins:
– The deadline for submission of your assignment is .
– For undergraduate students, you may do this assignment as a team of two stu- dents and hand in one submission per team.
– For postgraduate students, you have to do this assignment individually and make individual submissions.
– All implementations have to be done in C++.
– You need to submit your source code using the web submission system. You should
attach you and your partner’s name and student number in your submission.
– Late submissions will attract a penalty: the maximum mark you can obtain will be reduced by 25% per day (or part thereof) past the due date or any extension you are granted.
• Marking scheme:
– 20 marks for 10 randomly generated tests (2 marks per test).
If you have any questions, please send them to the student discussion forum. This way you can all help each other and everyone gets to see the answers.
The assignment
The aim of this assignment is to improve your learning experience in the process scheduling algorithms. You are required to design an online ticketing system for the Melbourne Cup held every year in Melbourne Arena. Specifically, there are 10,500 seats for public booking via this system for a single match. Figure 1 shows the distribution of the seat sections, where all the yellow and green areas are for public reservations.
In the system, all the customers are grouped into five priority classes (numbers), ranged from 1 to 5, according to their loyalty (accumulated points) to this ticketing system. A smaller priority number indicates a higher priority. All ticketing processes (purchase/cancellation) generated by a customer are assigned with the same priority number of that customer. To maximise the system service performance, you are required to implement a scheduling algorithm using multi- level queue strategy with two queues: a high priority Queue 1 and a low priority Queue 2, where Queue 1 has absolute priority over Queue 2. In other words, customers in Queue 2 will only be processed if there is no customer in Queue 1.
A new customer on arrival will be put on Queue 1 if his/her priority number ≤ 3 or Queue 2 otherwise.
Detailed actions in these queues are described below:
1
Figure 1: Hisense Arena
Queue 1: This is the High-Priority queue (priority number ≤ 3). Customers in this
queue are treated in the way of combined Highest Priority First (HPF) and Weighted Round
Robin (WRR — definition see below) on priority as follows: Select the highest priority customer
(customers with the same priority are processed in their arrival order), and then process this
customer’s request for a ticket quota of N = weighted time quantum tickets non-preemptively, then 5 time units
move this customer to the end of (the priority-subqueue of) Queue 1, where • weighted time quantum = (10 − customer’s priority number) × 10
• 1 ticket costs 5 time units to process.
• customer’s priority number is the customer’s current priority number.
Weighted Round Robin (WRR): Given n processes P1,P2,…,Pn, where process Pi has a weight wi (1 ≤ i ≤ n), WRR allocates Pi a weighted time quantum of wi time units (i.e.
w
weight is (10 − customer’s priority number), and customer’s priority number is the customer’s current priority number.
a share of n i CPU time). In this assignment, to simplify the implementation, a process’ i=1 wi
2
For example, The priority number 2 in this queue is decreased by one run of this customer, i.e. once booked the first N = (10−2)×10 = 16 tickets its priority will become 3 and its ticket
5
quota for the next run will book N = (10−3)×10 = 14 (instead of 16 in its first run). 5
Customers of the same priority are processed in their arrival order. The priority of a customer in this queue is decreased by one every 2 runs of this customer, i.e. when a customer has been processed 2 times under its current priority, its priority number will be increased by 1. When a customer’s priority number reaches 4, it will be demoted from Queue 1 to Queue 2.
Queue 2 : This is the Low-Priority queue (priority number ≥ 4). Customers in this queue are handled in Shortest Remaining Time First. That is, select the customer with shortest remaining time, regardless of their priorities, and process this customer’s request preemptively (to any a new arrival or promoted customer in Queue 1).
Customers of the same job length (time units) are processed in their priority number in- creasing order. Customers of the same job length and priority are processed in their arrival order.
Note: once a running customer X in this queue is interrupted by a promoted or new arriving customer in Queue 1, customer X will quit execution immediately and the new customer in Queue 1 will get the CPU to run.
Aging mechanism
Because Queue 1 has priority over Queue 2, and the HPF strategy is applied, starvation may occur. That is, some customers in Queue 2 may never get to run because there is always a job with higher priority in Queue 1. To address the starvation issue, you must implement the following aging mechanism that dynamically promotes a customer of Queue 2 to Queue 1 when he/she experiences starvation.
The priority of a customer in this queue (Queue 2) is increased by one every 100 time units run of other customers since the last run of this customer. That is, if a customer has waited 100 time units of other customers since his/her last run, his/her priority number will be decreased by one. In this way, the priority number of each customer in Queue 2 decreases gradually in proportion to the waiting time since the last run. When a customer’s priority number reaches 3, it will be promoted from Queue 2 to Queue 1.
Note:
• Order of simultaneous arrivals of same-priority customers:
For customers arriving at the same time with the same priority to both queues, we assume
their arrival order follows the increasing order of their customer IDs.
• Order of new arrivals, preemption and promotion
• For Queue 1, there may be three types of customers with the same priority arriving simul- taneously at the end of (the priority-subqueue of) Queue 1 — a new arrival customer A to Queue 1, a customer B with this priority of Queue 1 moved to the end of Queue 1 (by Weighted-Round-Robin), and a promoted customer C from Queue 2 to Queue 1. In this case, their execution) order in Queue 1 will be Queue1 → A → B → C
• For Queue 2, there may be two types of customers arriving simultaneously to Queue 2 — a new arrival customer A to Queue 2 and a demoted customer B from Queue 1, their relative arrival order in Queue 2 will be Queue2 → B → A.
Test Input
We have N customers, Each customer is identified by a line in the input file. The line describes the customer ID, arrival time(Can be divisible by 5), priority, age and the total tickets
3
required. For example a1 5 1 0 50 describes a customer a1 which arrived at time 5 with priority number 1 and age 0, and requires 50 tickets. One ticket processing consumes one time unit.
Output
The output includes N+1 lines. it provides information of each customer execution. First line is string include ”name arrival end ready running waiting”. The next N lines (sorted by termination time) should report the customer ID, arrival and termination times, ready time (the first time the system processes his/her request) and durations of running and waiting.
Web-submission instructions
• First, type the following command, all on one line (replacing xxxxxxx with your student ID):
svn mkdir –parents -m ”OS” https://version-control.adelaide.edu.au/svn/axxxxxxx/2020/s2/os/assignment1
• Then, check out this directory and add your files:
svn co https://version-control.adelaide.edu.au/svn/axxxxxxx/2020/s2/os/assignment1 cd assignment1
svn add TicketBooker.cpp
svn add StudentFile1.cpp
svn add StudentFile2.cpp
···
svn commit -m ”assignment1 solution”
• Next, go to the web submission system at: https://cs.adelaide.edu.au/services/websubmission/
Navigate to 2020, Semester 2, Operating Systems, Assignment 1. Then, click Tab ”Make Submission” for this assignment and indicate that you agree to the declaration. The automark script will then check whether your code compiles. You can make as many resubmissions as you like. If your final solution does not compile you won’t get any marks for this solution.
• We will test your codes by the following Linux commands: g++ TicketBooker.cpp -o TicketBooker
./TicketBooker input.txt>output.txt
• For undergraduate student write students name and ID on File Header comments of Tick- etBooker.cpp
4