Very short WUs these times. Am i the only one?
29 Nov 2006 20:54:53 UTC
Topic 192133
(moderation:
In the latest week or so i have only expierenced very small WUs. I am almost done with one WU now, and it has taken no more than 45 mins. Furthermore, i get only around 12.36 points or so for each WU. Is this how it is suppose to be or is something wrong? I also had a lot of WUs recently that didn't run at all but was removed immidiately (or almost immidiately) - it says compute error when i go to user results.
"The world is a fine place and worth figthing for." (Ernest Hemmingway)
"Non progredi est regredi
Very short WUs these times. Am i the only one?
)
I've received short tasks lately also but I don't consider this a problem. And 12 points for short tasks and 112 points for long tasks is not mysterious is it? My particular machine takes 1.2 hours for short and 11.5 hours for long tasks (hyper-threaded). So the points awarded are commensurate with the run time. Is your question 'Why am I getting short tasks instead of long'?
It is all completely
)
It is all completely normal.
The WUs come in 2 sizes with the long ones taking about 10X the short. The credits are scaled the same.
[edit]Here's the message from Dr. Allen about this.
Kathryn :o)
Einstein@Home Moderator
RE: In the latest week or
)
Getting runs of long WUs and short WUs is normal. I'd be more concerned about the compute errors. The last time I had them, I solved the problem by resetting the project and cooling the computer by aiming a floor fan at it. There are several topics out there that detail other reasons for compute errors.
Anyone got links to those topics on compute errors?
RE: RE: In the latest
)
If I am reading the BOINC Wiki correctly, the "exit code 107" errors he was getting, indicate "Cannot connect to the FTP site". Not sure of the timing, but the "outage" Einstein had a while back, might be the reason.
Yeah, the error messages are
)
Yeah, the error messages are quite strange, however i don't get them anylonger. I haven't had short WUs ever i think. Looking at my desktop i have had a lot of short WUs and still have, but at my laptop i am currently running a long WU. But let's see if it persist on the desktop. In a way i like the long ones the best though i know the point ratio is the same :)
"The world is a fine place and worth figthing for." (Ernest Hemmingway)
"Non progredi est regredi
I still get the long WUs on
)
I still get the long WUs on my laptop and the short WUs on my desktop, no exceptions i think, and i don't think it is a coincident as it has been this way for the last couple of months. Is there an option or something i have clicked at since i get them this way?
"The world is a fine place and worth figthing for." (Ernest Hemmingway)
"Non progredi est regredi
After looking at your hosts,
)
After looking at your hosts, they are both eligible to receive long work datapacks.
That being said, keep in mind that as long as there are new results to be run from the datapack you have onboard you will run that work until it's gone. Depending on how "fresh" the datapack was when you got it and some other factors, it can take a long time to plow through it all. I have a couple of hosts which have been working on the same datapack for a several months.
Alinator
Never heard about datapacks,
)
Never heard about datapacks, but it does sound like the explanation in this case. Can i get an overview of the datapacks that i am currently running? If not, i guess i can live without it now that i know what causes my desktop to get short WUs while my laptop gets long WUs.
"The world is a fine place and worth figthing for." (Ernest Hemmingway)
"Non progredi est regredi
The datapack is the big 15MB
)
The datapack is the big 15MB file which has the same root name as the results you are currently running.
It's basically a chunk of the data from one of the interferometers which is further subdivided by the hosts running the individual frequency template WU's against it looking for signs of a gravity wave. The template frequency is the second field from the left in the filename.
As you may be aware a template frequency less than 400 MHz is classified as the "short" WU's and greater than that are classified as the "long" WU's.
I'm pretty sure Dr. Allen mentioned how many individual searches (WU's) are made on the datapacks, but I don't recall offhand. I seem to remember that it isn't a fixed value. You can get a rough idea about where you are in the datapack by looking at the sequence number (the fourth from left field in the result filename). Lower is closer to the end of the datapack run.
One other fact about datapacks to keep in mind is they are initially sent to a limited group of hosts to work on. If any of those "drop out" of the game, it eventually gets sent to a new host to take over that slot and help finish it. That's why you get the occaisional "oddball" with a different TF for a few results and then pick up a new "fresh" one which runs for a long time.
I've even had a couple of instances where I got just a single result from a datapack to crunch. I'm assuming this was because it was getting very "stale" and the Project servers decided they were getting tired of looking at it, and wanted it to go away finally. ;-)
HTH,
Alinator
Thx for the info.. Where do
)
Thx for the info.. Where do you see this 'oddball' and what is TF short of?
"The world is a fine place and worth figthing for." (Ernest Hemmingway)
"Non progredi est regredi