Midterm Examination

1. ( 15 points) Sun's NFS uses RPC as its communication paradigm. For the file operations open(), close(), read(), and write(), what kind of RPC semantics are needed and why?

2. ( 10 points) Concurrent updates to a shared data item may cause the "lost update" problem, where all updates get "overwritten" by the last one. Which file system(s) covered in class are prone to suffer from this problem? Why?

3. ( 25 points) Consider introducing the notion of time as one way to fix the lost update problem in the file system(s) identified above.
   • ( 10 points) Describe how you would solve the problem.

   • ( 15 points) Discuss the use of Lamport's logical clock and Jefferson's time warp mechanism in the context of solving the lost update problem. Illustrate your answer with examples of applications with different file access patterns.

4. ( 20 points) An alternative to cache consistency maintenance is to have a guarantee from the server that, for a pre-specified time period, clients that are currently accessing a file will be consulted before a client can write to that file. Compare this approach to callbacks and timestamp-based invalidation in terms of robustness, consistency guarantees, and client-server communication.

5. ( 30 points) Describe the places in the protocol and OS/application stack where security services can be provided. What are the tradeoffs involved? Can we add security to a system at a single layer so that it protects all higher layers? Why or why not?