Sekerob said:
A look reveals the 4691 was reported by olivier3191, and going by his signature he's crunching 12.8 days/day. If assumes 0.5 hours/task, this equals 7.49 days cache. While this definitely is on the high side, if WCG isn't going to stop users that's away in the week from running WCG, it's a completely legal cache-size. Note, the 12.8 days/day is probably higher, since the signature is only slowly increasing, while the 0.5 hours/task can be too fast.


RaymondFO said:
And based on my own estimates on 03.09.2012, there was less than 3 days left of work-distribution and distribution-frequency seemed to increase indicating could be only 1-2 days left with work. Also, going by my own estimates, I needed 9.89 cache-days more of work, so did a "panic-cache" and set cache-size to 9.9 days.

Roughly 24 hours later, armstrdj posted his 1st. estimate about available work, telling it was roughly 14 days left of work.

Based on this estimate my large cache wasn't neccessary. Also, with longer than my expected time to the finish, it gave me the opportunity to use my computers for other things and still reach my CFSW-goal. This non-WCG-usage is the main reason needed to abort roughly 1200 tasks from one computer that anyway would have hit their deadline.

By my current estimate, I'll hit sapphire in 8 days. This is a further slow-down from before.


BTW, going a little off-topic...
Doing my own calculations afterwards, as already mentioned 1200 tasks = 4.4 days cache, going by a speed of 0.53 hour/task, since by my valid results most is either 0.52 or 0.53 hours. Add to this that AFAIK peaked at roughly 2600 tasks on computer, something that equals 9.57 days cached at 0.53 hour/task.

9.57 days is less than 9.9 days, so going by this the computer was never over-cached. But, based on BoincTask, the computer one time had atleast 75 days cached and possibly some higher, since it's a hex-core this equals at least 12.5 days cache.

Even before increased the cache-size this computer, according to BoincTask, had too much work compared to specified cache-size, but despite of this the computer happily grabbed more work with increased cache-size, and according to BoincTask had atleast 12.5 days of work afterwards...

So, while I won't rule-out either a client-bug or a server-bug (or both) allowing too much work to be downloaded, atleast to me it can look like it's BoincTask that can't add-up how much work is cached...

To continue from my previous posting above here are some guidelines or "best practices" that I use for any project:

1) During the life of the project, incrementally add time to reach your project goal. Make your goal realistic and obtainable given your current output and hardware configuration;
2) Reasonably estimate your run times to determine what is needed;
3) Assume your cache is the additional time you need to reach your goal after the estimated WCG project completion announcement. Prior to this announcement Sekerob's charts estimates of potential completion times is in my opinion the "gold standard";
4) Add a small "fudge factor" in your calculations to allow for unforeseen events; and
5) Download what you need and crunch it;

An estimated 6 days full supply left, then set in a few days your Additional Buffer in the client prefs to 7 days, and you got your 12 days. This is assuming a 1 thread crunching device. If you got 2 cores/threads, set your buffer to just a few days and you got reserve to reach your gold goal, nail-biting easy (though crunching is not meant to be stress inducing ;>)

Good luck... and thanks for spreading all your crunching time at WCG.

duanebong, I certainly haven't heard of an extension - although I, too, am hoping for one so as to give myself a chance at reaching my main goal of Emerald for all active projects. At the moment, it looks as though I'll be stopping at Ruby for this one (I estimate that I'd need another ~9 weeks to reach Emerald after I've hit the 180 day mark).

According to my calculations I officially achieved Sapphire, so tomorrow I will release the last 110 WUs that I suspended as I was pretty sure I did not need them.