Please submit your response on Gradescope by Thursday, October 31, 2024, 4:30 PDT 🎃.
- The NFS paper makes a big deal out of the protocol being stateless. In a few sentences, sketch what a stateful version of the NFS system might look like, perhaps by explaining how the signature of a few of the protocol procedures would change (or what procedures you might add). Explain how a client program would use your protocol to do something, e.g., write to a file, and compare the difficulty of crash recovery during this operation in your stateful design vs. the stateless NFS system.
- Both NFS and Leases experienced issues when clocks were not synchronized between machines. Briefly describe those issues, one for NFS and one for Leases. Which was probably most sensitive to clock skew?
- Sketch how modern computers synchronize their clocks, e.g., with the Network Time Protocol. Be sure to describe briefly how they account for network latency. Also give an order-of-magnitude for how well this works, e.g., if you choose two random computers, are their clocks off by 1s? 1ms? 1ns? (These are not in the papers, you should search in the library or online — be sure to cite your sources!)
- Given the same clock skew (e.g., 1s), which would you expect to be more sensitive to this skew and why: computers/applications 20 years ago, or computers/applications today? If you can’t decide, or if it depends on the application, etc., give one reason for each. Explain your reasoning in a sentence or two.