CSE410 Autumn 2000 Final Exam Solutions

MULTIPLE CHOICE (2 points each)

Questions 1-20 are multiple choice.  For these questions, choose the one best answer.

1. Which of the following is the reason that the least recently used (LRU) algorithm is usually not used as a page replacement algorithm?

A. Other practical schemes such as MIN do a better job.

B. LRU requires knowledge of the future to work correctly.

C. LRU is too inefficient to implement in practice.

D. The Clock algorithm always outperforms LRU.

 

2. Which of the following are shared between threads in the same process?

A. registers

B. page table

C. stack

D. stack pointer

E. None of these are shared

 

3. The IP service model guarantees which of the following?

A. The contents of a packet will not be corrupted

B. That at least one copy of a packet will arrive at the destination

C. That no more than one copy of a packet will arrive at the destination

D. That packets sent in order will be delivered in order

E. None of the above

 

4. Using the SCAN (a.k.a. elevator) disk scheduling algorithm, determine how far the disk head moves servicing the following outstanding requests: 150, 19, 20, 900, 99. Assume the disk's head is currently over track 100 and moving toward smaller track numbers.  Also, assume that all of these track requests have already been received.

A. 962 tracks

B. 1064 tracks

C. 1681 tracks

D. 1863 tracks

E. None of these are true

 

5. If a process has allocated every 1024th virtual page (e.g. it has allocated virtual pages 0, 1024, 2048, 3072, 4096, 5120 ... 1024000), which one of the following page table schemes will use the LEAST amount of memory?

A. A flat page table

B. A two-level page table with 1024 first level entries

C. A two-level page table with 2048 first level entries

D. An inverted page table

E. Each of the above page table will use exactly the same amount of memory

6. Making no assumptions about the processes being scheduled, which of these scheduling algorithms will prevent starvation?

I. First Come First Served  (doesn’t work if a job runs forever—infinite loop)

II. Round Robin

III. Priority

 

A. I

B. II

C. I & III

D. I, II, & III

 

7. Consider a group of RAID4 disks. This group has four data disks and one parity disk. Which of the following are true?

I. Any data disk but NOT the parity disk can be reconstructed from the other four

II. The parity disk must be read whenever one of the data disks is read

III. The parity disk must be written whenever one of the data disks is written

 

A. III

B. I & III

C. II & III

D. I, II, & III

E. None of the above

 

8. Which of the following is true about two threads running in the same process?

A. One thread can both read and write another thread's registers

B. One thread can change the other thread's program counter

C. One thread can neither read nor write the other thread's stack

D. One thread can both read and write the other thread's stack (there is no address space protection between threads in the same process)

 

9.Which disk block allocation scheme will require the most I/O operations for random access to a large file?

A. Indexed allocation

B. Linked allocation

C. Contiguous allocation

D. I-node allocation

E. Each scheme requires approximately the same number of I/O operations

 

10. Which of the following does NOT solve deadlock?

A. Acquiring all resources before using any of them   (this is not acquiring all resources or no resources.  This doesn’t eliminate hold and wait)

B. Only acquiring resources in order based on a predetermined priority

C. Acquiring a maximum of one resource at a time (can’t have hold and wait if only have one resource)

D. If a resource is unavailable, releasing all resources and waiting for all required resources to become available

11. What is the primary reason that a translation lookaside buffer (TLB) is used?

A. A TLB ensures that a process does not access memory outside of its address space

B. A TLB makes translating virtual addresses to physical addresses faster

C. A TLB allows multiple processes to share the L1 cache

D. A TLB makes translating virtual addresses to physical addresses possible

E. None of the above

 

12. Which of the following is not a solution to thrashing?

A. Running fewer processes

B. Increasing the speed of the CPU

C. Increasing the size of physical memory

D. Rewriting programs to have better locality

E. These all solve the problem of thrashing

 

13. Which of the following is NOT a way to make file systems faster?

A. Put parts of a file on many different tracks so part of it can be accessed no matter where the disk head is

B. Cache frequently used blocks in memory so they can be accessed at memory speeds instead of disk speeds

C. Use a disk scheduling algorithm to minimize the distance between seeks

D. Put frequently used files or directories near the center of the disk so on average they won't be far from the read head

E. All of these will make a file system faster

 

14. Which of the following is true about base and bounds registers?

I. They offer protection between processes

II. They lead to internal fragmentation of physical memory

III. Once a process has been started at a given memory location, it cannot be moved to another location

 

A. I

B. II

C. I & II

D. I, II, & III

E. None of the above

 

15. Which of the following can I do without knowing your private key?

A. Pretend to send a private message on your behalf

B. Decrypt messages that were intended for only you

C. Send you a message that only you can read

D. Digitally sign a message on your behalf

E.  I must know your private key to perform all of these operations

 

16. Public/private key encryption is usually avoided because

A. It is much slower than symmetric key encryption

B. It is only useful when encrypting large amounts of data

C. It is not as secure as symmetric key encryption

D. It cannot be used unless two parties can exchange their private keys securely

E. None of the above

Suppose a system has only three physical pages. Given the following sequence of virtual page references, determine the number of page faults that are required. Initially, assume that the physical pages are not being used by any virtual page.

1 2 1 1 3 2 1 4 3 1 1 2 4 1 5 6 2 1

17. Using the first in first out (FIFO) replacement policy

A. 9

B. 10

C. 11

D. 12

E.  None of the above

 

18. Using the least recently used (LRU) replacement policy

A. 9

B. 10

C. 11

D. 12

E.  None of the above

 

 

19. Which of the following is NOT a use of TCP sequence numbers?

A. To ensure that data is delivered to the application in the order it was sent

B. To distinguish duplicate packets

C. To determine how much data is left to send

D. All of these are uses

 

20. Which of the following is NOT a property of Ethernet?

A. Nodes retransmit packets if there was a collision

B. Sending packets that are too big does NOT affect collision detection

C. Nodes retransmit packets if an acknowledgement is not received

D. After a collision, both nodes wait a random amount of time before retransmitting a packet

E. These are all properties of Ethernet

 

SHORT ANSWER

 

None of these questions requires a long answer.  Some require an answer of only a few words.  Others may require up to a few sentences.  The amount of space left to answer the question is not indicative of how long your answer should be.  Your answers should be specific.

 

21. Earlier in the quarter, we used the following equation to calculate the average memory access time:

AMAT  = L1HitProb * L1HitTime +   (1-L1HitProb)*( L2HitProb * L2HitTime + (1-L2HitProb)*L2MissTime )

In this equation, either an L1 cache miss or an L2 cache miss can lead to an increased memory access time.

When processes use virtual addresses (i.e. processes do not refer to memory using actual physical addresses), what additional event can increase the average memory access time? Be specific. (3 points)

            TLB miss

When using virtual memory (i.e. infrequently used portions of memory are moved to the page file), what additional event can increase the average memory access time? Be specific.  (3 points)

            Page fault

 

22. Consider a process that has two threads. One thread puts items on the tail of a queue, and the other thread removes items from the head of the queue. Britney argues that you don't need synchronization because the two threads are accessing separate ends of the queue. Her friend, Christina disagrees. Who is right? Briefly explain you answer. (7 points)

Christina is right.  Consider the case that the queue has only zero or one elements.  Both threads might be accessing the same element.

 

 

 

23. Assume two threads are executing the following code segment simultaneously. Can there be deadlock? Why or why not? (6 points)

Lock1->Acquire();

...

Lock2->Acquire();

...

Lock1->Release();

...

Lock2->Release();

 

There cannot be deadlock because there is no circular wait. 

 

24. With a 2GByte disk, Microsoft's FAT16 file system uses disk blocks that are 64Kbytes. With the same disk, their FAT32 file system uses disk blocks that are 4KBytes.

Give one specific reason in favor of using 64KByte blocks. (3 points)

Accessing files might be faster because they are stored contiguously.

 

Give one specific reason in favor of using 4KByte blocks. (3 points)

There is less internal fragmentation than with 64Kbyte blocks.  That is, not as much space is wasted for each file.

 25. In the Internet, why is public/private key cryptography typically used instead of symmetric key cryptography? (5 points)

If symmetric key encryption was used, then the two parties must exchange the symmetric key secretly, which is usually impossible.  Public/private key encryption does not require a secret to be transferred between the two parties in order for them to communicate securely.
26. Below is an implementation of a BankAccount class. (8 points)

  class BankAccount {
  private:
    double balance;
    Mutex mutex;
   public:
    BankAccount() { 
      balance = 0.0;
    }
    double GetBalance() {
      return balance;
    }
    void DepositMoney( double amount ) {
      mutex.acquire(); 
      double newBalance = balance + amount;
      balance = newBalance;
      mutex.release();
    }
    void WithdrawMoney( double amount ) {
      mutex.acquire();
      double newBalance = balance - amount;
      balance = newBalance;
      mutex.release();
    }  
}; 

(Ignore the possibility that amount could be negative or that amount withdrawn could be greater than the balance.)

Assume that multiple threads could access a member of the BankAccount class at the same time. Assume a bank account’s balance is $100.  If one thread calls bankAccount-> DepositMoney(50.0) and one thread calls bankAccount->WithdrawMoney(75.0), what are all of the possible values for balance that could result after both of these threads execute? 

25, 75, 150

 

 

 

 

 

Now add synchronization to this class so that the balance that is stored is always correct.  You may change or add lines to the provided implementation. You may also rewrite the entire class definition if you desire.  You are not required to know the exact syntax for whatever synchronization primitive you choose to use.

(see code)