My WU's aren't sent to others?
30 Jul 2005 18:38:31 UTC
Topic 189639
(moderation:
I've been looking through my WU's, & I just realized that some of my WU's are just about expired, yet I am the only one that has that WU? What gives?
http://einsteinathome.org/workunit/1649117
http://einsteinathome.org/workunit/1649116
http://einsteinathome.org/workunit/1649119
My WU's aren't sent to others?
)
The way E@H sends out WUs is to send the large input file, then create WUs from that. This method can create issues such as you're seeing, because the scheduler needs to find hosts that have the same input file as you. It can't do this until that host contacts the scheduler. Which if it's a slow host, could take a week. You also need to take into account whether the host that may have the same input file is over committed, so it may not actually request any work on scheduler contact.
I'm sure there is some window of time when the scheduler decides it hasn't found anyone, and sends the input file to another host, but I have no idea what it is. This time frame might also be influenced if the scheduler knows there are hosts out there with the same file, but they haven't yet asked for work. If so, it might just wait indefinitely until they are ready to accept work, but that's just a guess.
RE: I've been looking
)
Don't worry -- you'll get credit for all of your work (assuming the results are valid, which I am sure they are). Although your WU are near, at, or past deadline, the other WU for other host machines may still be in progress.
Please see the front page FAQ for more details of how credits are granted. In your case, just be patient, wait, and keep crunching. Your credits are 'in the bank'.
Bruce
Director, Einstein@Home
RE: I'm sure there is some
)
More to the point, the sending of the input file to another host will only happen when
1) some other client asks for work, and
2) the scheduler has no results to match whichever file that client already has
whenever both of these occur, the scheduler has to send a new file to that client. Sometime soon it will be your file that gets sent, and then your wu will get sent out to that client as well, and you will start to get credit once they are returned.
Bruce says the credit is in the bank: maybe it is more like the credit is in the pipeline.
You are lucky: modem users will envy you because you will get more goes at this file than anyone else, a big saving on download time (if that matters to you)
~~gravywavy
RE: More to the point, the
)
I just joined, and the first work unit I received estimates 7:10:00 CPU time. When I check up on that work unit, I see mine is the fourth PC doing that work.
Is there an upper limit to how many PC's get the same work unit? If two or three PC's have the same work unit already, can't it send a new work unit? I mean, isn't there a vast amount of work to be done?
I understand everybody gets credit, but I want to know that my PC is being useful in some way, and it seems overly redundant being the fourth PC assigned to a single work unit. It's not about the credit.
Shouldn't there be an upper limit of 2-3 per WU set?
If it is that there are too few work units, there is no need for me to volunteer my PC time. Is there a way to get a different work unit?
A couple of links from the
)
A couple of links from the wiki that might help explain things for you.
http://boinc-doc.net/boinc-wiki/index.php?title=Quorum_of_Results
http://boinc-doc.net/boinc-wiki/index.php?title=Quorum_Size
http://boinc-doc.net/boinc-wiki/index.php?title=Quorum_Rule
http://boinc-doc.net/boinc-wiki/index.php?search=quorum&go=Go
Jim
Thank you. I realise the
)
Thank you. I realise the fourth PC is to save time because one of the first three could have a client problem.
I chose 'einstein' because it isn't a simulation, and I am not convinced that simulations are worth very much.
RE: Thank you. I realise
)
It's not just to save time but also to save resources. When you look at how many computers get added to the project but then end up returning little or no work, you will realise that someone has probably done a "simulation" :) (yeah, one of those low value thingys) to show that the strategy of sending 4 gives the best overall efficient use of server resources.
Next time you go flying as a passenger in a 747 and you land safely at your destination, spare a thought for the pilot who has trained for many hours on a simulator. Think of how much more your ticket would have cost and how much less safer it would have been if all aircrew had to do all their training on the real thing. Or next time you fill up the tank in your vehicle, consider how that stuff originally came from a hole in the ground where the people who operated the machinery that drilled that hole safely and efficiently, used simulation as an important part of their training. I help train drilling rig crews. I use a simulator. Every aspect of modern day life benefits from widespread use of simulation. Without it we would be in the technological dark ages.
Boinc projects that use simulation include CPDN and LHC. The Seti project analyses real data. Which of these would you consider to be of most "value"? If we found an intelligent signal even relatively close at hand just on the other side of our own galaxy, a mere 60,000 light years away, sure it'd be very interesting. But what practical use (once the hoohaa had died down) would it really have? How would it protect our grandchildren's grandchildren from the impending ravages of climate change, for example? Don't you think that the particular simulation in CPDN may help produce the necessary changes in attitude needed right now to protect the future of this planet?
Sorry, this wasn't meant to be a rant but rather an impassioned defense of the benefits of simulation of anything that might impact on our everyday lives.
Cheers,
Gary.
RE: RE: I chose
)
I don't see the value of SETI, make no mistake. Simulation is not bad per se, I can fully appreciate the value of training simulators. However, this is not about training.
It seems to me that a simulation only serves to illustrate how certain preconditions affect the whole picture. Showing that the climate will get hotter doesn't seem like a great achievement to me. It's been said and done.
Quantifying exactly how it might become doesn't seem too valuable. We know the climate is worsening. We don't need a simulation to tell us that.
RE: We know the climate is
)
Actually, we don't. Not really. One of the problems with weather is that we do not know that much about the WHY. We can look at things in the fossil record, tree rings and so forth, but is the warming trend because of humans? Or is it just a variation because it is a variation.
We cannot answer this question cleanly becasue we have no weather simulation that makes sense. This is one of the reasons that the 5 day forecasts are so poor. We just cannot run the models that far out and get answers that work. Is it because the model is poor, or because we don't use a fine enough granualrity, or something else...
That is why we need to do simulations. Which ones work, and which ones don't maybe we can, by looking at them, tell why. So, we have 3 projects doing simulations and two looking at real measurement data. And I personally find both of them interesting.