Advanced Operating Systems : Exam Guide. Question/Material Distribution Guide. Exam Type Questions

Transcription

1 Advanced Operating Systems : Exam Guide There will be a two-hour exam, and students will be asked to attempt four questions from a choice of five. In designing the five questions I ll refer to the following rough distribution guide. Question/Material Distribution Guide 1. Shells, shell scripts, fork, pipe, wait, basic Linux commands. (Based on Lectures 1,2,3.) 2. Forking, pthreads, OpenMPI, concurrent programming. (Based on Lectures 2,3,4,10,11.) 3. Concurrent programming, deadlock, 2-phase-locking. (Based on Lectures 5,6,7.) 4. Briefly describe four ways of dealing with deadlock. 5. Networks, Amdahl s law and its application, parallel algorithms. (Based on Lectures 12,13,14,15.) 6. Parallel algorithms and speedup. (Based on lectures 8,9,16,17,18) Exam Type Questions (These questions are only intended to indicate the kind of question that could come up on the exam, and the list of questions is certainly not an exhaustive list of possibilities.) 1. (a) Write a C shell script that starts the gap computer algebra system in a new terminal. The new terminal should use black font on a white background. Assume that gap has been installed in the usual way on your computer. (b) Suppose you have saved your script in the file script.sh. Explain carefully how you could run the script. (c) Explain how you could include the sentence Click here to start gap into a LaTeX document so that your shell script could be started from the resulting pdf document by clicking on the underlined word. 2. Write a C shell script that outputs to the terminal the current date, time and number of users currently logged in. 3. What would the following Linux script output to the terminal? $ find. -name "*.cpp" -print wc -l Briefly explain (using the words fork, pipe, exec ) how the script would perform its task. 4. What is a critical section in code? Explain the three properties that any solution to the Critical Secion Problem should guarantee. 5. Describe Dekker s algorithm for solving the Critical Section Problem. 6. What is deadlock? Give four conditions that must hold in order for deadlock to occur. 7. Which one of the following resource allocation graphs depicts deadlock? Explain your answer. R1 R2 R3 R1 R2 R3 P1 P2 P3 P4 P1 P2 P3 P4 1

2 8. Explain carefully what the following code does. Your explanation should contain a detail description of how fork is used in the code. #include <stdio.h> #include <sys/types.h> #include <unistd.h> #include <sys/ipc.h> #include <sys/shm.h> #include <stdlib.h> main(int argc, char **argv) int n, i; double d, s, x, pi; int shmid, iproc; pid_t pid; double *shared; n = atoi(argv[1]); d = 1.0/n; shmid = shmget(ipc_private, sizeof(double), (IPC_CREAT 0600)); shared = shmat(shmid, 0, 0); shmctl(shmid, IPC_RMID, 0); iproc = 0; if ((pid = fork()) == -1) fprintf(stderr, "The fork failed!\n"); else if (pid!= 0) iproc = 1 ; s = 0.0; for (i=iproc+1; i<=n; i+=2) x = (i-0.5)*d; s += 4.0/(1.0+x*x); pi = d*s; if (pid == 0) *shared = pi; else wait(0); pi = pi + *shared; printf("pi=%.15f\n", pi); (Q2) 9. Explain carefully what the following code does. Your explanation should contain a detail description of how pthreads are used here. #include <stdio.h> #include <pthread.h> #include <stdlib.h> 2

3 int n, num_threads; double d, pi; pthread_mutex_t reduction_mutex; pthread_t *tid; void *PIworker(void *arg) int i, myid; double s, x, mypi; myid = *(int *)arg; s = 0.0; for (i=myid+1; i<=n; i+=num_threads) x = (i-0.5)*d; s += 4.0/(1.0+x*x); mypi = d*s; pthread_mutex_lock(&reduction_mutex); pi += mypi; pthread_mutex_unlock(&reduction_mutex); pthread_ main(int argc, char **argv) int i; int *id; n = atoi(argv[1]); num_threads = atoi(argv[2]); d = 1.0/n; pi = 0.0; id = (int *) calloc(n,sizeof(int)); tid = (pthread_t *) calloc(num_threads, sizeof(pthread_t)); if(pthread_mutex_init(&reduction_mutex,null)) fprintf(stderr, "Cannot init lock\n"); ; for (i=0; i<num_threads; i++) id[i] = i; if(pthread_create(&tid[i],null, PIworker,(void *)&id[i])) exit(1); ; ; for (i=0; i<num_threads; i++) pthread_join(tid[i],null); printf("pi=%.125f\n", pi); 10. Compare forking, pthreads and MPI as tools for concurrent computing. 3

4 11. What is a shell? Describe one major difference between the implementation of the ls command and the cd command in the C shell. 12. Name and describe the six basic MPI functions used in six function MPI code. (Q2) 13. Explain what the following MPI C code does... (Q2) 14. Define the following parameters for a network of processors: number of nodes, number of links, network latency, diameter, bisection width. Briefly explain the practical significance of each parameter. 15. For each of the following networks determine: the number of links, the diameter, and the bisection width. (a) A completely connected network with n nodes. (b) A line network with n nodes. (c) A ring network with n nodes. (d) A k k mesh network with n = k 2 nodes. (e) A k k torus network with n = k 2 nodes. (f) A hypercube network with n = 2 k nodes. (g) A complete binary tree of height h. (Also determine the number of nodes in this case.) 16. Show how a ring network with 9 nodes can be perfectly embedded into a suitable k k torus network. Also, find the minimum dilation with which a 10 node ring network can be embedded into a 4 4 torus network. 17. Explain how a 4 4 torus network can be perfectly embedded into a hypercube network with n = 2 k for any k What is the minimum k for which a complete binary tree with 31 nodes can be embedded into k k torus network. Describe such an embedding and calculate its dilation. 19. Give a precise statement and proof of Amdahl s Law. Suppose that you are applying for funding to build a network to make a specific computation, and at most 80% of the computation lends itself to parallelization. How many computers would you apply for? 20. Describe bucket sort. Prove that, if the data is evenly distributed in a given range, then one could expect almost a linear speedup when bucket sort is implemented on m processors. 21. Describe a pipeline sorting algorithm for sorting a list of length n on n processors. Illustrate how the algorithm would sort the letters SLIGO into alphabetical order. Calculate the maximum speedup of the algorithm. 22. An unsorted list has been divided and distributed over the nodes in a hypercube network of 8 nodes in a fair but arbitrary manner. Describe how a parallel QuickSort algorithm can be used to compute the sorted list. Illustrate the algorithm for the list 1,,16 distributed as follows. ( Q5) 4

5 7,11 2,12 1,16 4,13 6,10 8,9 5,15 3, Consider the system of two data-base transactions T 1 = P b P a V b P c V a V c T 2 = P a P b V a V b expressed using Djikstra s P-V notation. Represent the system graphically, and then determine: (i) any regions of deadlock, (ii) whether the system is safe, and (iii) whether the system is two-phase locked (2PL). (Q3) 24. Consider the system of two data-base transactions T 1 = P a V a P b V b T 2 = P a V a P b P c V b V c. Represent the system graphically, and then determine: (i) any regions of deadlock, (ii) whether the system is safe, and (iii) whether the system is 2PL. 25. Describe the proof that two-phase locked systems are safe. 26. Given a set of rectangles A i, design a parallel program to position the rectangles in a large rectangle, subject to the following conditions: i) the individual rectangles must not overlap; ii) all the smaller rectangles must be included in the large one; iii) the area of the larger one is to be a minimum. 5