CS 417 Exam 1

Fall 2014

    Part I – 20 Points

  1. 4 POINTS
    Is the statement below true or false? Explain your answer. You should use a reliable transport protocol such as TCP for clock synchronization.
  2. 4 POINTS
    Explain the benefit of multi-canonical marshaling.
  3. 4 POINTS
    Why is using only reference-count based distributed garbage collection a problem?
  4. 4 POINTS
    Cristian's clock synchronization algorithm assumes symmetric network latency: the latency of sending a message is expected to be the same as that of receiving a response. This is not true in all networks. How would you modify the formula for setting the time if the downlink is 4 times faster than the uplink?
  5. 4 POINTS

    You have a set of servers that manage shopping cart data. Clients may send updated carts to any server and servers propagate these updates to other members of the group. Unfortunately, the network sometimes gets partitioned temporarily so only a subset of servers get updates. You need to reconcile the shopping cart for a user. The vector timestamp is a set of values associated with processors P0, P1, ... Pn. Items in the cart are a subset of {A, B, ... Z}. Causal updates should overwrite previous contents. Concurrent updates merge with previous contents. What is the final set of items in the user's shopping cart when reconciling the following timestamps?

    	Clock = { P0: 2 } Value = { A, C }
    	Clock = { P0: 1, P1: 2 } Value = { A, B, Q }
    	Clock = { P0:1, P1: 2, P2: 1 } Value = { A, E, H }
    	Clock = { P0: 1, P1: 2, P2: 2 } Value = { A, K }

    For each statement, select the most appropriate answer. I will give you credit for two incorrect answers.

  7. Compared to home snooping, source snooping:
    (a) Uses less bandwidth and has a lower latency.
    (b) Uses more bandwidth and has a higher latency.
    (c) Uses more bandwidth but has a lower latency.
    (d) Uses less bandwidth but has a higher latency.
  8. The most efficient use of a network is:
    (a) Time division multiplexing (TDM).
    (b) Frequency division multiplexing (FDM).
    (c) A token passing protocol.
    (d) Random access.
  9. Which of the following was NOT a design goal of the Internet?
    (a) Provide reliable communication.
    (b) Support the interconnection of different physical networks.
    (c) Use routers to store and forward packets.
    (d) Have decentralized control of the network.
  10. A port number in IP protocols is used in the:
    (a) Data Link layer.
    (b) Network layer.
    (c) Transport layer.
    (d) Session layer.
  11. The TCP layer does not deal with:
    (a) Reliable data transfer.
    (b) In-order data transfer.
    (c) Flow control.
    (d) Secure data delivery.
  12. Sockets are said to be compatible with files because:
    (a) They can be named with textual names.
    (b) You can use open and close system calls.
    (c) You can use read and write system calls.
    (d) All of the above.
  13. In RPC, the client stub:
    (a) Implements the remote function on the client but uses remote access for all data in that function.
    (b) Is an automatically-generated placeholder function where the programmer fills in the logic.
    (c) Is a wrapper that loads the server function.
    (d) Presents the functional interface of the remote function and sends a request to the server.
  14. In RPC, marshaling refers to:
    (a) Serializing the data of an object for transmission.
    (b) Sending a message to invoke a remote function.
    (c) Looking up the address of a server.
    (d) Packaging the code for a function so it can be sent and executed remotely.
  15. An interface definition language is used to:
    (a) Enable the generation of stub functions.
    (b) Implement the server-side remote functions.
    (c) Access remote objects.
    (d) Define interfaces for all local and remote classes in a program.
  16. A linear compensation function:
    (a) Makes a clock run faster or slower by a constant amount.
    (b) Converts a variable-frequency clock to a fixed-frequency clock.
    (c) Sets the system time forward or back by a fixed amount to set the time to the desired value.
    (d) Generates periodic timer interrupts to allow an operating system to keep track of time.
  17. A client requests a timestamp from the server at 3:52:20.200. The server response is received at 3:52:20.600. The response contains a server timestamp of 3:52:20.400. The client should set its time to:
    (a) 3:52:20.200
    (b) 3:52:20.400
    (c) 3:52:20.600
    (d) 3:52:20.800
  18. A client synchronizes from a server. It sends a request at 3:52:20.200. The time between a request and a response is 52?ms. The best-case time is 40 ms. What is the error of the synchronization?
    (a) ±6 ms
    (b) ±12 ms
    (c) ±20 ms
    (d) ±106 ms
  19. If event A has a Lamport timestamp of 2 and event B has a Lamport timestamp of 3, you are certain that:
    (a) A happened before B if the events took place on the same process.
    (b) A happened before B regardless of the processes on which A and B took place.
    (c) A and B are concurrent.
    (d) B happened before A, regardless of the processes on which A and B took place.
  20. Which event is concurrent with the vector timestamp { 5, 2, 1, 4 }?
    (a) { 6, 3, 2, 5 }
    (b) { 5, 2, 1, 5 }
    (c) { 4, 2, 0, 4 }
    (d) { 5, 1, 2, 4 }
  21. Which of these multicasts can be implemented with a group-wide sequence number server?
    (a) Global time ordered multicasts
    (b) Total ordered multicasts
    (c) Partial ordered multicasts
    (d) Unordered multicasts
  22. A hold-back queue:
    (a) Allows a sender to queue messages if the receiver is too congested to receive them.
    (b) Holds messages at the sender until they are acknowledged in case the sender needs to retransmit them.
    (c) Enables the receiver to receive messages at a constant rate.
    (d) Allows a receiver to switch the order of received messages before giving them to the application.
  23. A precedence vector in a message is used by a receiver to:
    (a) Ensure that all messages are received in the same order by all group members.
    (b) Create timestamps of when each message has been received.
    (c) Identify whether all previous causally-related messages have been received.
    (d) Create a log of received messages in the order they were received..
  24. IGMP, the Internet Group Management Protocol, is used to:
    (a) Build a global list of hosts that are subscribed to a given multicast group.
    (b) Tell a router that a connected host is interested in receiving a multicast group.
    (c) Enable system administrators to manage multicast subscriptions.
    (d) Allow a multicast sender to discover and contact a multicast receiver.
  25. Which mutual exclusion algorithm requires, on average, the fewest messages in a group of 100 machines?
    (a) Centralized.
    (b) Token-ring.
    (c) Lamport's.
    (d) Ricart & Agrawala.
  26. The Ricart & Agrawala mutual exclusion algorithm:
    (a) Does not depend on time stamps in messages while Lamport?s does.
    (b) Cannot handle the case where two or more processes request the same resource at the same time.
    (c) Does not require a process to send messages to the entire group while Lamport?s does.
    (d) Requires fewer messages than Lamport?s algorithm.
  27. The Chang & Roberts ring election algorithm optimizes the ring election algorithm by:
    (a) Contacting only higher-numbered processes.
    (b) Multicasting the election request instead of passing messages in a ring.
    (c) Choosing a leader as messages are sent instead of sending a growing list of candidates.
    (d) Having a node always discard an election message that came from a process with a smaller ID.
  28. A practical way to handle network partitioning is to:
    (a) Have each process start an election.
    (b) Elect a leader only if a majority of nodes can be reached.
    (c) Run a normal election and multicast the result to the subset of processes that can be reached.
    (d) Divide and conquer: elect multiple coordinators.