Other projects aren't that picky...

Ulrich Metzner
Ulrich Metzner
Joined: 22 Jan 05
Posts: 113
Credit: 936,370
RAC: 0
Topic 188831

See here

As stated, other projects, e. g. predictor, aren't that picky on late results. I was on holidays and the report deadline of 7 days is a joke, c'mon.

Aloha, Uli

Shaktai
Shaktai
Joined: 8 Nov 04
Posts: 183
Credit: 426,451
RAC: 0

Other projects aren't that picky...

> See href="https://einsteinathome.org/%3Ca%20href%3D"http://einsteinathome.org/task/2343337">http://einsteinathome.org/task/2343337">here[/url]
>
> As stated, other projects, e. g. predictor, aren't that picky on late results.
> I was on holidays and the report deadline of 7 days is a joke, c'mon.

Not a joke, but current project need. BTW, Predictor also has a 7 day deadline, so Einstein isn't the only one. This is scientific research. I imagine the deadlines are based upon research requirements and project goals. The deadline issue keeps coming up on these projects, and as more projects adopt BOINC, we will likely see even greater diversity in deadline requirements. Just going to have to get used to it. Every project is different. Every research requirement is different. Each participant must decide if they can meet the project requirements.

Mike
Mike
Joined: 20 Feb 05
Posts: 109
Credit: 944,815
RAC: 0

Hi Because of the sience i

Hi

Because of the sience i think a project should not loose any data.
Everybody would switch einstein off when going to holiday.

greetz Mike

Todd Wright
Todd Wright
Joined: 6 Mar 05
Posts: 13
Credit: 39,039
RAC: 0

> Not a joke, but current

Message 9874 in response to message 9872


> Not a joke, but current project need. BTW, Predictor also has a 7 day
> deadline, so Einstein isn't the only one.

Yes but an Einstein WU takes about 7 times longer to process than a Predictor one - at least on my machines.

Seti for example has a deadlne of 14 days which is "Max days between connects" (10) + 4 for good measure. The E@H deadline is set at 30% less than Max days. (or 7). You dont have to be a scientist to see that its too short - It is not even compatible with the user configurable options.

People have suggested to me that I change the Max days to 0.1 which is ridiculous and defeats the purpose of that settting, which is to cache WU to allow for the occasional project downtime.

As for Mike's comment:

> Everybody would switch einstein off when going to holiday.

What for? I have several machines that run 24/7. I just spent 2 weeks in Japan, and they ran perfectly while I was away (With the exception of E@H). Should I shut them down? Should I bring down the web servers and mail servers and other things that they run, just because I am not nearby? I dont think so.

The botttom line is that the combination of E@H's deadline being too low, with the BOINC client scheduler requesting too much work, is causing me (and other people) to seriously consider detaching my hosts from this project.

I have it running on 2 right now (I have already detached a slower machine) but if even my main crunchers cant complete their allocated WU before the deadline, it is not worth the effort.

Mike
Mike
Joined: 20 Feb 05
Posts: 109
Credit: 944,815
RAC: 0

Hi As i wrote in an other

Hi

As i wrote in an other thread i have running three machines with seti and einstein.
Two of them has internet connection once a week so i have to set connect to 10 days that they have enough work.
So its nearly impossible for that two to reach the deadline from einstein.
from 3 results 2 will not getting credits so its a question of time i will suspend einstein on this machines because the deadline.

greetz Mike

Ulrich Metzner
Ulrich Metzner
Joined: 22 Jan 05
Posts: 113
Credit: 936,370
RAC: 0

Even Pirates@home has

Even Pirates@home has extended their former deadline of 1 (!) hour to 24 hours, regarding their ~10 minutes wu's. If you're doing multi projects (the main new feature of BOINC!) then it is sometimes mandatory to extend the deadline. And as stated before BOINC allows a setting of 10 days. So at least 14 days deadline seems reasonable to me.
BTW: I'm not worried about the not given credit but my wasted enthusiasm ;)

Aloha, Uli

The worm that turned
The worm that turned
Joined: 11 Feb 05
Posts: 9
Credit: 559,721
RAC: 0

The deadline for units is

The deadline for units is what it is.
no one is forced to process Einstein units.
And no one will be missed if they go, there are plenty more signing up each day.

FalconFly
FalconFly
Joined: 16 Feb 05
Posts: 191
Credit: 15,650,710
RAC: 0

Back in the days where there

Message 9878 in response to message 9877

Back in the days where there were only few Projects, I would agree that 7 days would be perfectly sufficient.

But consider this :

The more Projects are active, the more each User can attach to his machine (which the whole point and greatness of BOINC).

When one day there are 10 Active Projects, an equal share would only yield 2.4 hours of CPU time per day for each Project, including Einstein.
(all this in a perfectly normal situation and keep in mind there is a whole fleet of Projects either considering BOINC or already running internal Alpha testing on it; it also additionally assumes the machine is running 24/7 which is usually not the case)

With that situation coming up, even fast machines will one day easily scratch the 7 day limit, caching a 2nd WorkUnit not even being considered...

IMHO 14 days is more "future-proof", and the Server-side Infrastructure and deadlines should sooner of later reflect the changing situation of increasingly available BOINC Projects.

Relying solely on Users that run Einstein dedicated or only few Projects is therefor the wrong approach.

Rule No.1 for Distributed Computing :
Leave noone behind!
(especially if the "fix" is so easy and the Computing share is rapidly and increasingly being diffused/spread among countless attractive Projects)

The former, unchallenged No.1 (SETI) has already begun to lose its previous grand share to all the other Projects.
What Einstein (or any other Project) darely needs (moreso in the future) is to do everything needed to attract and maintain as much of the available Computing Assets as possible.
If that means either increasing the deadline or reducing the average runtime per WorkUnit accordingly, it should be done better today than tomorrow.

PS.
Several other Projects invest an increasing amount of effort to optimize their Clients for maximum performance to get the most out of their limited share. They've basically realized the Situation and took the required measures to counteract their resources slowly bleeding off due to above reasons.

Mike
Mike
Joined: 20 Feb 05
Posts: 109
Credit: 944,815
RAC: 0

Hi Just my 2c. For

Hi

Just my 2c.

For sience its important to have so much volunteers as possible.
The slower machines did the same work than the faster ones but hasn´t the same chance to get credits for.
My brother lost last week 3 units, what is 18 houres of crunching because five minutes reported to late and he will loose 3 again this week.
The einstein units takes nearly the double time to process to one seti unit but has the half time to send back.
I think einstein could have much more users.

@theworm
i know that for sure.

greetz Mike

Shaktai
Shaktai
Joined: 8 Nov 04
Posts: 183
Credit: 426,451
RAC: 0

The deadline is determined by

The deadline is determined by the project need. If the Science needs 7 days for maximum effeciency and results, then that is what it is. Science is improved by good science, not necessarily more users. The project team must designate the need for the project.

I have all 7 of my computers working on Einstein and Predictor, the two projects that have the shortest deadlines. The deadlines are not a problem for me. I can see where they might be for some other folks. The participants must make a decision based upon their capabilities. Rather then trying to have every computer do everything, identify what your setup can handle and then participate more fully in those projects that meet that criteria. The smart cruncher, focuses his computer resources where he/she can do the greatest good, rather then just scattering everything about to satisfy ego.

More users does not necessarily equate to better science. It is not a numbers game, it is the results that are important. For now, it would appear that the most important thing for the project, is to have the results back quickly. That importance of timely feedback outweighs the desire for more participants. In time that could change, as long as it does not compromise the scientific effort. The participants must decide if they can support the project need.

*To change the world for the better, start by changing yourself for the better.*

Mike
Mike
Joined: 20 Feb 05
Posts: 109
Credit: 944,815
RAC: 0

Hi Sorry, but this is a

Hi

Sorry, but this is a short time thinking.
More results is more sience, simply matematics.
I dont think that more 2 or 3 days of a longer deadline makes a limit of the project.
I don´t have a problem with that, but think it could more done with more flexibility.

greetz Mike

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.