Does a fully utilized disk imply the scheduler is doing a good job interleaving i/o and CPU bound jobs?

If we can store persistent data on a local disk, we can also store it on a
remote one. In this case, if you cache file blocks rather than reading them from the
remote disk on every access you notice that decreasing the block size results in less
network traffic when you evict a cached block. What's a likely reason for this?

Would you expect an extent-based file system to have more or less space
overhead than FFS as the disk becomes increasingly full? (Recall: an extent-based file
system simply adds a counter to each disk pointer, which tracks the number of blocks
the pointer refers to.)


Give a reasonable disadvantage to using hierarchy in the systems covered during lecture.


On an exokernel operating system, does the code that decides how many disk blocks to prefetch on file read reside in the library operating system or in the exokernel itself?

What speed trick can we do to optimize context switching for exiting or newly created threads?

When a phone is disconnected, phone companies try to keep a FIFO list for re assignment to minimize the chance that the new owner will get calls intended for the old recipient. Briefly explain what similar problem TCP sequence numbers are used to solve (and how).

While we primarily use caches for 'speed, how can they increase reliability for name translation in a distributed system? (E.g., to cache the binding between DNS name and IP address, between IP address and Ethernet address, between DNS suffixes and name servers, etc.)