Книга: Distributed operating systems
PROBLEMS
PROBLEMS
1. A university is installing DCE on all computers on campus. Suggest at least two ways of dividing the machines into cells.
2. DCE threads can be in one of four states, as described in the text. In two if these states, ready and waiting, the thread is not running. What is the difference between these states?
3. In DCE, some data structures are process wide, and others are per thread. Do you think that the UNIX environment variables should be process wide, or should every thread have its own environment? Defend your viewpoint.
4. A condition variable in DCE is always associated with a mutex. Why?
5. A programmer has just written a piece of multithreaded code that uses a private data structure. The data structure may not be accessed by more than one thread at a time. Which kind of mutex should be used to protect it?
6. Name two plausible attributes that the template for a thread might contain.
7. Each IDL file contains a unique number (e.g., produced by the uuidgen program). What is the value of having this number?
8. Why does IDL require the programmer to specify which parameters are input and which are output?
9. Why are RPC endpoints assigned dynamically by the RPC daemon instead of statically? A static assignment would surely be simpler.
10. A DCE programmer needs to time a benchmark program. It calls the local clock time procedure (on its own machine) and learns that the time is 15:30:00.0000. Then it starts the benchmark immediately. When the benchmark finishes, it calls the time procedure again and learns that the time is now 15:35:00.0000. Since it used the same clock, on the same machine, for both measurements, is it safe to conclude that the benchmark ran for 5 minutes (±0.1 msec)? Defend your answer.
11. In a DCE system, the uncertainty in the clock is ± 5 sec. If source file has time interval 14:20:30 to 14:20:40 and its binary file has time 14:20:32 to 14:20:42, what should make do if called at 14:20:38? Suppose that make takes 1 sec to do its job. What happens if make is called a second time a few minutes later?
12. A time clerk asks four time servers for the current time. The responses are
16:00:10.075±0.003,
16:00:10.085±0.003,
16:00:10.069±0.007,
16:00:10.070±0.010,
What time does the clerk set its clock to?
13. Can DTS run without a UTC source connected to any of the machines in the system? What are the consequences of doing this?
14. Why do distinguished names have to be unique, but not RDNs?
15. Give the X.500 name for the sales department at Kodak in Rochester, NY.
16. What is the function of the CDS clerk? Would it have been possible to have designed DCE without the CDS clerks? What would the consequences have been?
17. CDS is a replicated data base system. What happens if two processes try simultaneously to change the same item in different replicas to different values?
18. If DCE did not support the Internet naming system, which parts of Fig. 10-24 could be dispensed with?
19. During the authentication sequence, a client first acquires a ticket and later a PAC. Since both of them contain the user's ID, why go to the trouble of acquiring a PAC once the necessary ticket has been obtained?
20. What is the difference, if any, between the following messages:
{{message }KA}KC and
{{message }KC}KA.
21. The authentication protocol described in the text allows an intruder to send arbitrarily many messages to the authentication server in an attempt to get a reply containing an initial session key. What weakness does this observation introduce into the system? (Hint: You may assume that all messages contain a timestamp and other information, so it is easy to tell a valid mes sage from random junk.)
22. The text discussed only the case of security within a single cell. Propose a way to do authentication for RPC when the client is in the local cell and server is in a remote cell.
23. Tokens can expire in DFS. Does this require synchronized clocks using DFS?
24. An advantage of DFS over NFS is that DFS preserves single-system file semantics. Name a disadvantage.
25. Why must a clone of a fileset be in the same disk partition?
- Appendix B. Common problems and questions
- Problems loading modules
- mIRC DCC problems
- Troubleshooting Runlevel Problems
- Troubleshooting Connection Problems
- Avoiding Printer Support Problems
- Delegation Problems
- Reverse Lookup Problems
- Troubleshooting Problems in Zone Files
- Runtime Errors, Boot Loader Problems, and Kernel Oops
- Preparing for Potential Hardware Problems
- How to read