On Rosetta, I have set my target CPU time to 12hrs. Is this good or bad? I'm about a third of the way through 8 WUs.
Since some of the tasks really should be returned within 18 hours, 12 hours is on the large side... Basically, you've got 3 options:
1: Only crunch for the credit, and don't care if the results isn't usable for CASP9...
2: Decrease target run-time, probably 6 hours would be a good target.
3: Use <report_results_immediately>, not normally recommended since adds unneccessary extra server-load, but during this race this would be an option.
For both #2 and #3, cache-size shouldn't be more than 0.1 days. So small cache-size isn't optimal if you're also taking part in Collatz, but one way around this problem is to set Rosetta to "no new work", fill-up with Collatz-work, for so decreasing cache-size again and enabling Rosetta work-requests again.
The reason for cache-size and target run-time:
0.1 days = 2.4 hours.
2.4 hours + 12 hours = 14.4 hours turnaround-time.
But... if supposes single-core and not <report_results_immediately> (multi-core is more unpredictable), cached work won't drop below 0.1 days before 9.6 hours into next task, meaning turnaround-time is.. 14.4 hours + 9.6 hours = 24 hours > 18 hours.
If 6 hours on the other hand:
0.1 days + 6 hours = 8.4 hours.
8.4 hours + 3.6 hours = 12 hours turnaround < 18 hours.
Even if task takes a couple hours longer than the target, something that it can do, you should normally manage returning it before the 18-hour deadline.
(you can actually get the same turnaround-time with 0.2 days cache-setting, but with 0.3 the turnaround-time will increase).
