1. What properties should the entity identifiers/names in distributed systems have? What are the differences between names and addresses? 2. Compare the differences between name servers deployed at the global layer, the administrational layer, and the managerial layer of a large-scale name space.

3. Discuss the advantages and disadvantages of iterative and recursive name resolution, respectively.

4. Explain the working of the network time protocol (NTP) and the Berkeley Algorithm, respectively. Discuss the differences between the two algorithms.

5. Identify all the causally related events and concurrent events in the following figure

6. When use Lamport’s logical clocks to derive a total order of events in distributed systems, why each message needs to be acknowledged?

7. What are the limitations of Lamport’s logical clocks, and how does the vector clock algorithm solve the problems?
8. Explain why mutual exclusion is needed when machines in a distributed system concurrently access shared resources. Compare the advantages and disadvantages of token- and permission-based approaches for mutual exclusion.

9. Discuss the tradeoff between large and small consistency units (conit) in keeping data consistent.

10. Discuss the differences of sequential, causal consistency and their relation to the eventual consistency.

11. Compared the advantages and disadvantages of busy-waiting, blocking synchronizations and monitor-based synchronization.

12. Explain the working of the MESI cache coherence protocol.

13. Compare the differences of the four eventual consistency models: monotonic-read, monotonic-write, read-your-write, and write-follow-read.

14. What are the requirements for building a dependable system?

15. Discuss the advantages and disadvantages of the flat and hierarchical structures in a process group, respectively.

16. Explain why it is much more difficult to realize fault tolerance in distributed systems with arbitrary node failures (e. g., node exhibiting byzantine behaviors) than systems with other types of failures, e. g., crash, omission, and timing failures.

17. Describe the requirements for atomic multicast

18. Discuss how the two-phase distributed commit protocol handles coordinator and participant failures and what the limitation of the two-phase commit is.

19. Discuss the different types of security threats and the objectives of a secure system.

20. Compare secret key and public key based cryptography and explain how the two schemes work.

21. Quantitatively compare the cost of three different distributed locking protocols. i. e., decentralized, distributed, and token ring, in a distributed system with 2000 nodes. Assume that sending and receiving a message in the distributed system takes 2ms. The frequency of accessing the critical section is one access per second and on average there are three concurrent access requests at a time. Calculate the average cost of accessing the critical section (i. e., the time cost) in the three protocols. You can assume that the time spent in the critical section is zero and all cost comes from messaging.

22. Consider the following situation, in which machine A sends a clock synchronization request to time server B. The request has a timestamp of T1 = 1:52:19.200 (reads as one o’clock, fifty-two minutes, nineteen seconds and 200 milliseconds). Time server B receives the request at T2 and sends a response with its local timestamp at T3 = 1:52:19.500. The response is received at A at time T4 = 1:52:19.600. The network time protocol (NTP) is used for synchronization. Assume that the latency of the time server to respond to a clock synchronization request is 100 milliseconds, what is the offset of A’s clock relative to that of the time server and what timestamp should A set its clock to? Show your work and explain the practical meaning of the formula you use.

23. Consider the following communications between processes P0, P1, P2. Answer the questions below.

What are the vector timestamps of the six remaining events?

24. Give examples (using diagrams) of the following client-centric consistency models: monotonic reads, monotonic writes, read-your-writes, and writes-follow-reads.