Faults, victims, and thrashing

Computing · Computer Systems · ages 20-21

Name ______________________ Date ____________

No frame is free when a fault arrives. What happens?
- The machine melts
- The fault is ignored
- Another page is evicted as victim
What triggers a page fault?
- A page that is not resident is referenced
- A program prints text
- A disk spins quietly
What does thrashing look like from outside?
- A silent idle disk
- Constant disk activity with tiny progress
- Faster programs
Same string under first in first out. How many faults, and why equal here?
- 9, because FIFO double counts
- 0, because frames are many
- 4, because the repeat hits under both policies
Three frames, refs 1, 2, 3, 1, 4. Count faults under least recently used.
Answer: ______________
The optimal policy evicts the page used farthest in the future.
Circle one: True False
A kernel could implement the optimal policy by logging past references.
Circle one: True False
Sam claims least recently used always faults less than first in first out. What is wrong?
- Nothing, it always wins
- The taught string ties at 4 faults each, so always is false
- Policies never fault

Answer key

For grown-ups. Fold this page away before handing over the rest.

Another page is evicted as victim · A victim must leave before the newcomer loads.
A page that is not resident is referenced · Missing residency traps to the kernel for a fetch.
Constant disk activity with tiny progress · Page moving replaces real work.
4, because the repeat hits under both policies · The early repeat hits either way, and only the tail evicts.
4 · Loads 1, 2, 3 fault, 1 hits, and 4 faults: 4 total.
True · Farthest-next-use is the perfect victim choice.
False · Past logs are not future knowledge, so it stays unimplementable.
The taught string ties at 4 faults each, so always is false · One tie disproves always, as the example shows.