Nobody would deny that OS3 has caused tears before bedtime on several occasions. Fortunately, as we were relieved to find out, the degree exam is notable for the absence of C coding exercises.
Question 1 was about (non-Real-Time) priority-based scheduling. I went into the exam swearing that I wouldn't take on a question on that topic, but the question was attractive, compared to Q3. Some tricky definitions did arise, however, with "Decay Function" and "Usage Measure" destined to become the "Track Buffer" of 2004. The talk of Linux's "Multi-Level Feedback Scheduler" was a little confusing, as Stallings appears to give a different definition from the notes - presumably, however, there is some cross-over.
Question 2 was the question for which almost everyone was waiting. Part (a) was a 22-mark behemoth on the subject of virtual memory (the classic "Load a value" story), for which - I should hope - everybody prepared, just in case. A cheeky little 3-mark rider was tacked on the end, asking about approximate timings. Given the juxtaposition of disk access and electronic calculation, one presumes only the former (at six orders of magnitude greater) needed to be considered.
Question 3 followed the same pattern as Algorithmics' question 2: an enticing first-half (about file management on disk, familiar from past papers), followed by a surprising second-half. The single-sheet lecture notes on Operating System Structure took care of a bookwork question (three definitions), followed by an R/A-question about privileged mode, which I didn't fancy at all. Although this question tends to include components from the end of the course, I'm not sure that many would expect this part to come up (no precedent, certainly, not that I'm suggesting it shouldn't have been included), and I doubt it was a strong area for most.
All that remains to be said is that Peter Dickman runs a tight examination hall.