Fast computer with 20 pending results
28 Feb 2005 15:44:46 UTC
Topic 188134
(moderation:
Hi again.
A user over at the Bad Astonomy BB team has had 21 completed, successful WUs on an Athlon 2700+; and has received credit for only the first one. His WUs take an average of 5.8 hours. (21000 s)
He is really stressing out over failing to get credit.
I've counselled patience, you have to wait 7 days, slower computers are probably working on the verification, etc., but I fear he may lose patience and go back to distributed.net.
He has a second box, running at 803 Mhz with almost as many credits as his Athlon. (each reporting 1 WU.)
It is true that his oldest WU is from Feb 23; which is still within the 7, but yet is almost 5 days. None of the other members of our group has had to wait anything like this long.
He does seem extraordinarily snakebit: that oldest pending WU has one "No Reply" and the fourth computer is fast, but is apparently only on sporadically, and so has no chance of confirming before deadline.
Similar things seem to be going on with the next few WUs in order of age.
Is there anything else I can say to him, or any chance something more is going on than just a bad roll of the dice?
We have several other members (some high profile :)) with 3.2+ Ghz boxes that are just coming into their reporting windows, and I really would be hard pressed if the same thing happens for them.
To put it starkly, how does one respond to new users who ask:
Why are fast computers slow to get credit, and slow computers fast to get credit?
[I understand the reason for the second part: the probablity is high that the other voting computers are faster than you, so you're the last to report & bingo credit is givem.]
I know this oversimplifies the situation, and the FAQ comes close, but talking directly to this point might be good. It seems a shame that we might lose folks just because their computers are too fast!
Rereading, this sounds like a terrible rant, but I truly don't mean it that way -- more a cry from the heart or whatever, and so I'll leave it as is and hope you'll understand.
Ken
Fast computer with 20 pending results
)
Ken,
I understand. Just today, I finally passed the threshold where I have more credits awarded than I do in pending status. Not much, but just more.
For most of the work units I've crunched, I am not the fastest. For the ones I have pending, many WUs have been there 5-6 days. One of the credit groups I received came from WUs where the two slowest finally timed out.
I spent the first 5 or 6 days feeling like I was doing all of this work for naught, that all of my results would be thrown away. I know the frustration. I finally reconciled myself that this project is a team effort and that crunching a WU was a mini-team effort. If I was the fastest (or second fastest), I could take some small amount of pride in that I had finished what my mini-team was working on and had gone on to new WUs while they "lagged" behind.
I understood from some post somewhere that WUs were assigned to crunchers based on similar results. I'm not sure I believe this. The guy who consistently beats me for first place is twice as fast as I am. He's on an Athlon 64; I'm on a pokey P4, 2 gig. I'm wondering what kind of computer the guys who are DAYS behind me are running...
If he's in this for short term reward, as in racking credits like a slot machine spins, he'll never be happy. If he's in it for the good feeling that comes from being even a small part of the advancement of science, he might. Trust me. For every major, heart fulfilling, internally rewarding night of observatory time, there are literally hundreds of cold, lonely, disappointing nights of boredom.
There are two secrets to life: 1) Don't tell everything you know...
> Why are fast computers slow
)
> Why are fast computers slow to get credit, and slow computers fast to get
> credit?
>
> [I understand the reason for the second part: the probablity is high that the
> other voting computers are faster than you, so you're the last to report &
> bingo credit is givem.]
>
> I know this oversimplifies the situation, and the FAQ comes close, but
> talking directly to this point might be good. It seems a shame that we might
> lose folks just because their computers are too fast!
I'll add something to the FAQ specifically to address this. In the way of motivation, you might tell the user that when the slow machines finally DO get around to repeating the work that he has done, his Recent Average Credit (RAC) scores may really shoot through the roof, moving him or her high up in the list.
Bottom line: counsel patience. If the user is returning successful results, they should get lots of credit. But it may take a week or two to start arriving.
Cheers,
Bruce
Director, Einstein@Home
Bruce, Thanks very much
)
Bruce,
Thanks very much for your post; I'll pass it on immediately. I know it's unscientific to believe in Bruce Allen karma, but he's just been credited with another WU. :)
Star Charter,
What a beautiful post!! Thanks so much, I know it will mean a lot to the folks over there.
In the few days I've been fielding questions for this new team, I've been so immersed in credit issues, there's been almost no time to look up, at what we're all actually doing.
I'm trying not to bawl all over the keyboard here :), but really: thanks.
Ken
>Why are fast computers slow
)
>Why are fast computers slow to get credit, and slow computers fast to get
>credit?
>[I understand the reason for the second part: the probablity is high that the
>other voting computers are faster than you, so you're the last to report &
>bingo credit is givem.]
You gave the simple answer in your original post Ken: If you are running a fast computer the odds favor that the other computers assigned that workunit will not be done when you are and you will have to wait. If you are running a slower computer the odds favor you being the last one to report and you will get almost instant credit.
After you have been crunching for 2 to 3 times the normal deadline for a project the number of pending should stabalize. Then you start getting credit granted at about the same rate you generate pending credit.
BOINC WIKI
BOINCing since 2002/12/8
I joined a few days ago. I've
)
I joined a few days ago. I've downloaded about 24 WU and uploaded about 16 completed ones. What bothers me is that only one of those WUs have been sent to any other computers. Why is this?
http://einsteinathome.org/account/tasks
> I joined a few days ago.
)
> I joined a few days ago. I've downloaded about 24 WU and uploaded about 16
> completed ones. What bothers me is that only one of those WUs have been sent
> to any other computers. Why is this?
>
> http://einsteinathome.org/account/tasks
>
I had the same problem early on. There is some discussion in some thread on this, but the bottom line is just wait a few days to a week or so. Most computers are given a week or so for the work and all the WU may not go out for 3 or 4 days. Face it not everyone that gets a WU completes it and sends it back. Eventually they will be sent again till the work is done and credits are given.
Bottom line, it can take up
)
Bottom line, it can take up to and sometimes more than two weeks to get credit..............
http://boinc.berkeley.edu/participate.php
Link to Unofficial Wiki for BOINC, by Paul and Friends
Getting "started" takes a
)
Getting "started" takes a bit, but once your backlog of pending WU's start clearing, the credits really start rolling in and it's pretty consistent from then on it seems to me. It is just in the beginning you have to develop a backlog of pending work units, so patience in the beginning is required.
Sometimes if you have a quick
)
Sometimes if you have a quick computer, you are givin WU's that haven't been dished out to anyone else, you just have to wait for the rest of computers to catch up.
> I joined a few days ago.
)
> I joined a few days ago. I've downloaded about 24 WU and uploaded about 16
> completed ones. What bothers me is that only one of those WUs have been sent
> to any other computers. Why is this?
>
> http://einsteinathome.org/account/tasks
>
Steve,
Why don't you start a thread on the "Problems" board to make this more visible? It seems very weird that you have so many pending results that have not been sent to any other computers. You need to get someone like BM to look into this.
Cheers,
Gary.
Cheers,
Gary.