More Stars ???

log in

Advanced search

Message boards : Number crunching : More Stars ???

Previous · 1 · 2 · 3 · 4 · 5 · Next
Author Message
Profile Coleslaw
         
Avatar
Send message
Joined: 11 Apr 10
Posts: 182
Credit: 8,165,591
RAC: 2,123
Total hours: 19,356,045
Message 1465 - Posted: 30 Jul 2013, 22:09:37 UTC - in response to Message 1463.

I am not one of those with 1000's of computers, but I do run about a dozen older systems in my home. The key is to find resources everywhere possible. (With permission or rights to do so of course) Over the years I have had several borged machines come and go that have added probably over double the contribution I would have done on my own. Anyone could do that if they just made the effort. I have even exchanged free labor for the promise to run BOINC when idle. I never install it without permission though. I am even donating computers to friends/family/needy that have it loaded on there. They are usually happy to leave it on there in exchange for the free PC.

Now, if you talk to the hard core DC'ing community, they tend to not have thousands of systems. (Those are reserved for the lucky few and organizations that participate.) Many of the hard core DC'ers purchase high end equipment with multiple GPU's and multiple processors. I know there are individuals on my team that have multiple quad socket setups. Most of them run FAH instead of BOINC. When each processor has 16 cores, you can get some pretty good production out with a single system. A few of my teammates run these on BOINC part of the time. Just one of those servers outdo my entire house for less electricity. If it weren't for me using mostly recycled equipment, it would also be financially smarter in the long term. However, my wife would never approve of the noise.

So, for the average Joe, the current badge system probably is enough as is. But there are many who need the challenge. I don't care what time measurement they go with as long as it includes everyone. I also would hate to see badges disappear even temporarily because of the complaining that would follow. People don't take that well. I for one would prefer to see someone come up with a badge system that you can have degrading badges. That way to maintain it, you have to keep active. Since using points only works when comparing one project, I don't think basing such a badge on RAC would work here. So, the only way that idea would work with run time is by removing apps as they are ended at projects. Again, this kind of badge would only have a small following and would need to have been instituted from the get go. Just a random rambling thought. :)
____________

Profile skgiven
       
Avatar
Send message
Joined: 7 Sep 10
Posts: 453
Credit: 945,109
RAC: 0
Total hours: 2,101,570
Message 1466 - Posted: 30 Jul 2013, 23:01:50 UTC - in response to Message 1465.
Last modified: 30 Jul 2013, 23:27:32 UTC

While I agree that any badge system should appeal to everyone, and I like a challenge, I would't like to only have a RAC based badge system; I concentrate on a few projects and for the rest I'm quite nomadic. One problem with a would be RAC system is that it would favor continuous even contribution over intermittent, but some projects/apps/research is inherently intermittent, so such a system would have the undesired effect of being project-biased.
Another problem is that if someone did the same hours for one project/app in a (one off) week as someone else did in a year (but evenly) the former would have zero credit and the latter's credit would have risen to and remained at a constant level.
If someone jumps around between 20projects they could have little or no credit. That would defeat the point of having a credit system, and would certainly not be for everyone.
They have something akin to that at FoldIt! though its position/points based. While I use to participate there I no longer do, and partially because of that system. I did like getting referenced though.

Utilizing what you can and optimizing resources and performance/Watt is part of the fun, but it's very project specific. Not that I consider myself part of the HC DC community.

PS. There are 5 parts to the ring in the badge,

(Thanks to Sellyme)

If those 5 parts were used as well as the 9 stars that could cover 14*20apps (280) - just saying...

Sellyme
   
Send message
Joined: 21 Jul 13
Posts: 69
Credit: 691,597
RAC: 0
Total hours: 717,632
Message 1468 - Posted: 31 Jul 2013, 0:56:13 UTC - in response to Message 1465.

Many of the hard core DC'ers purchase high end equipment with multiple GPU's and multiple processors. I know there are individuals on my team that have multiple quad socket setups. Most of them run FAH instead of BOINC. When each processor has 16 cores, you can get some pretty good production out with a single system.


Yep, three users on my team run multi-processor setups (quad Opteron 6272s, and two dual-Opteron 6278s with two spare CPU sockets for upgrades), and a fourth just should have his quad 6178 system arrive today. That being said, systems like that are (as you've addressed), VASTLY better on Folding@Home than BOINC, due to the multi-threaded processing capabilities of F@H, and the quick-return bonus. The quad-6272 system ran MilkyWay@Home for a few days to test it out, but it was decided that it was pretty much just a waste of resources and that BOINC is better to run on low-to-mid range computers, with high-end computers on F@H.

Of course, this isn't the case for everyone, but as you've addressed, it is for the majority.

So, for the average Joe, the current badge system probably is enough as is. But there are many who need the challenge. I don't care what time measurement they go with as long as it includes everyone. I also would hate to see badges disappear even temporarily because of the complaining that would follow. People don't take that well.


Do you believe that my proposed system does not include everyone? Seeing as it reduced the requirements for every current step, as well as adding a particularly easy badge for newbies, I feel like it's significantly more inclusive than the current one. Similarly, for those who need the challenge... Well, get back to me when you have over 5000 years of run time and I'll happily agree to extend the badges again.

I for one would prefer to see someone come up with a badge system that you can have degrading badges. That way to maintain it, you have to keep active. Since using points only works when comparing one project, I don't think basing such a badge on RAC would work here. So, the only way that idea would work with run time is by removing apps as they are ended at projects. Again, this kind of badge would only have a small following and would need to have been instituted from the get go. Just a random rambling thought. :)


While this is a nice idea in theory, in practice it's just irritating and confusing. I've experimented with custom-made stats based on RAC, and it's just really unwieldy.


PS. There are 5 parts to the ring in the badge,

http://i.imgur.com/Km0cEkO.png (Thanks to Sellyme)

If those 5 parts were used as well as the 9 stars that could cover 14*20apps (280) - just saying...
http://wuprop.boinc-af.org/img/test_new_badge/badge9st.png


I would heavily recommend against splitting up the ring, that would just be way too small and finnicky (and, let's admit it, one-fifth of the ring would be a pretty shitty badge).

9 stars would cover 180 apps, which will keep us covered for quite a long time. As I explained in my first post here, any more than 120 apps for a badge right now is a pretty bad idea. Maybe if/when it's necessary to increase that, we could use a different shape?

Profile skgiven
       
Avatar
Send message
Joined: 7 Sep 10
Posts: 453
Credit: 945,109
RAC: 0
Total hours: 2,101,570
Message 1472 - Posted: 31 Jul 2013, 10:41:44 UTC - in response to Message 1468.

9 stars will not keep all of us covered for even a very short time!

Zombie67 has already crunched for 192 apps, would have 7 stars, and is 5apps short of an 8th star. Lots of apps come and go, as do projects. Obviously Z is a trailblazer on this front, but others will follow imminently.

At present there are 100 Boinc users who have 10K Boinc Credits for 75 Projects or more. Around 300 people have crunched for 50 or more projects, and 1000 people have crunched for 30 or more projects.
Note that many projects have lots of different apps. For example, at WCG I've crunched for 17 projects, some of which have more than one app, some apps changed and these projects had Beta apps. We're talking ~40apps over the last 5years. Many other projects have had more than 10apps in that time.

Since Boinc started there has been at least 135 different projects never mind apps.
Within a year or two it's inevitable that 100's of crunchers will have crunched several hundred different apps.

WUProp is helping more and more of the lesser projects by making crunchers aware of them, so it's also likely that the number of people who crunch for many projects will increase rapidly.

This is all moot anyway.

Sellyme
   
Send message
Joined: 21 Jul 13
Posts: 69
Credit: 691,597
RAC: 0
Total hours: 717,632
Message 1473 - Posted: 31 Jul 2013, 14:44:55 UTC - in response to Message 1472.

Yes, zombie is probably only 6 months to a year away from having 9 possible stars, but remember, he would then have to get them all up to 5 years of runtime. That would last quite a bit longer.

Profile skgiven
       
Avatar
Send message
Joined: 7 Sep 10
Posts: 453
Credit: 945,109
RAC: 0
Total hours: 2,101,570
Message 1474 - Posted: 31 Jul 2013, 15:02:39 UTC - in response to Message 1473.

We don't "have to get them all up to 5 years of runtime" - we can run for as long as we choose, if we have the opportunity; lots of apps won't be around long enough. Plenty of Beta apps will run once, and you may only get 1WU, never mind 100 or 5000h worth! Some Beta apps will reach 100h, some 500h but many will fall way short of 100h:

    World Community Grid Beta - Say No to Schistosoma 1.63 0.00
    physics@home coherence film sintering nano 0.37 0.00 (and I think this isn't even a Beta app)
    Constellation Scilab Alpha Test 0.07 0.00

I probably have 10apps under 100h that I will never get another WU for, and probably another 10apps that I have over 100h (and some over 5000h) that I will never get another WU for, simply because the app has been deprecated.
____________
=sticky tape=

Profile Coleslaw
         
Avatar
Send message
Joined: 11 Apr 10
Posts: 182
Credit: 8,165,591
RAC: 2,123
Total hours: 19,356,045
Message 1475 - Posted: 31 Jul 2013, 16:04:19 UTC

Sellyme, I'm not saying your proposal is a bad one. I'm not really targeting your idea specifically at all. I think the badge system needs to incorporate both higher run time limits and accommodate for a larger number of apps. What Skgiven points out about abandoned apps and alpha/beta is certainly a concern. I would like a way to filter out such apps just for convenience if nothing else. At least if they did not make the first badge cut anyways. That way I don't have to read through them every time I'm looking through the active apps. The problem would be for the project to know the apps were finished. I don't know if there would be an easy way to do that.
____________

Sellyme
   
Send message
Joined: 21 Jul 13
Posts: 69
Credit: 691,597
RAC: 0
Total hours: 717,632
Message 1480 - Posted: 1 Aug 2013, 4:14:08 UTC - in response to Message 1474.

Some Beta apps will reach 100h, some 500h but many will fall way short of 100h:

I probably have 10apps under 100h that I will never get another WU for, and probably another 10apps that I have over 100h (and some over 5000h) that I will never get another WU for, simply because the app has been deprecated.


And that's exactly why we shouldn't have too many apps on the badges: because it's just frustrating when they're literally impossible to achieve. When they ARE actually possible to achieve, then we can expand them again.

That way I don't have to read through them every time I'm looking through the active apps. The problem would be for the project to know the apps were finished. I don't know if there would be an easy way to do that.


Personally I think it would be a great idea to have a list somewhere of apps that have returned data in the last week, and that shouldn't be too difficult.

Profile skgiven
       
Avatar
Send message
Joined: 7 Sep 10
Posts: 453
Credit: 945,109
RAC: 0
Total hours: 2,101,570
Message 1482 - Posted: 1 Aug 2013, 12:00:30 UTC - in response to Message 1480.

Some Beta apps will reach 100h, some 500h but many will fall way short of 100h:

I probably have 10apps under 100h that I will never get another WU for, and probably another 10apps that I have over 100h (and some over 5000h) that I will never get another WU for, simply because the app has been deprecated.


And that's exactly why we shouldn't have too many apps on the badges: because it's just frustrating when they're literally impossible to achieve. When they ARE actually possible to achieve, then we can expand them again.

Impossible for who?
From the outset it would be impossible to ascertain who can attain which levels for different apps. Z might be able to get 50K, Coleslaw might be able to reach 10K, I might reach 2.5K and others won't make it to 100h. Even during the apps lifetime you would not know how long is left or how many hours each cruncher can contribute.

Who's going to decide when an app is deprecated?
Then there is the issue of apps being intermittent, or just not being sent to some systems due to WU requirements (some projects need 8GB RAM, others want high bandwidth...). Your system might not get/start work because of what else it's crunching; if you are crunching for one project that uses lots of RAM then another project won't start because the system does not have enough resources. Your system could have a task for days or weeks on but not run it. So is the app active for you, for everyone... It's too much work and far too complex to implement anything like that.

Apps not lasting long should be considered a different problem from the number of stars or their level. You don't need to tackle this issue via the badge levels or star count. Doing so is probably the wrong answer.
Perhaps apps that don't reach the first badge level could be separated (though they are already at the bottom of the list) and their average used for another star once the total is 2000h. This might still require that they be defined as inactive though.

That way I don't have to read through them every time I'm looking through the active apps. The problem would be for the project to know the apps were finished. I don't know if there would be an easy way to do that.


Personally I think it would be a great idea to have a list somewhere of apps that have returned data in the last week, and that shouldn't be too difficult.

I don't think there is an easy way but WUProp@home - Results tell us what's been returned recently and we can also see this by going to http://wuprop.boinc-af.org/hosts_user.php, clicking Activity and then stipulating the time period. Maybe the Results could be used to specify a projects status (active or inactive) and then our apps list could group accordingly, however this might change things in a way some people won't like.

I would prefer if we could just have the row highlighted a different way for active or inactive work, and keep the order according to the hours completed.

Perhaps if we returned work within 24h the app details could be highlighted by one color, if work was returned by others within 24h it could be a different color, if no work was returned in 24h the row could be highlighted as a third color, and a fourth color if no work was returned by anyone in a week. Alternatively the Pending (hours) could just be highlighted. For example,
    World Community Grid The Clean Energy Project - Phase 2 2,634.13 25.22
    World Community Grid The Clean Energy Project - Phase 2 2,634.13 0.00
    World Community Grid The Clean Energy Project - Phase 2 2,634.13 0.00
    World Community Grid The Clean Energy Project - Phase 2 2,634.13 0.00

Here Green means you crunched WU's in the last 24h, amber means others have but you have not, red means no work was completed by anyone for that app in 24h and black means no work was returned by anyone in a week or more.

Sellyme
   
Send message
Joined: 21 Jul 13
Posts: 69
Credit: 691,597
RAC: 0
Total hours: 717,632
Message 1484 - Posted: 1 Aug 2013, 15:36:27 UTC - in response to Message 1482.
Last modified: 1 Aug 2013, 15:37:50 UTC


Impossible for who?

For anyone. Correct me if I'm wrong, but I do not believe that there are 180 non-deprecated applications running, much less 180 stable ones.

[quote]Who's going to decide when an app is deprecated?


Hopefully the owners of the project. Having a "deprecated" flag for specific apps or projects as part of BOINC's system is sorely needed.

Then there is the issue of apps being intermittent, or just not being sent to some systems due to WU requirements (some projects need 8GB RAM, others want high bandwidth...). Your system might not get/start work because of what else it's crunching; if you are crunching for one project that uses lots of RAM then another project won't start because the system does not have enough resources. Your system could have a task for days or weeks on but not run it. So is the app active for you, for everyone... It's too much work and far too complex to implement anything like that.


It's possible to buy more RAM or get a server with high bandwidth. It's not (logistically) possible to magically produce another 20-40 apps to get to a badge that necessitates that.

Apps not lasting long should be considered a different problem from the number of stars or their level. You don't need to tackle this issue via the badge levels or star count. Doing so is probably the wrong answer.
Perhaps apps that don't reach the first badge level could be separated (though they are already at the bottom of the list) and their average used for another star once the total is 2000h. This might still require that they be defined as inactive though.


The system is complicated enough as it is, I feel like making it more convoluted would be a bad idea.

Perhaps if we returned work within 24h the app details could be highlighted by one color, if work was returned by others within 24h it could be a different color, if no work was returned in 24h the row could be highlighted as a third color, and a fourth color if no work was returned by anyone in a week.


I feel like 24 hours is a bit short of a timeframe, how about a week for the first three colours, and a month for the fourth? Sure, for users with dozens or hundreds of computers, you can reliably see what tasks you're active on from 24 hours, but for users running one or two computers, it's difficult to return work in more than 6-8 apps within a 24 hour period, even if you have 20+ active.

Profile KPX
       
Send message
Joined: 6 Apr 10
Posts: 45
Credit: 872,006
RAC: 0
Total hours: 1,547,205
Message 1511 - Posted: 9 Aug 2013, 14:55:46 UTC - in response to Message 1345.
Last modified: 9 Aug 2013, 15:01:21 UTC

Still looks to me as the best proposal. And if few crunchers conquer it, they can get a special badge (a big blue star, perhaps)... :-)
Edit: I looked at the stats of the largest teams. People, who now have 5 stars, count in single digits. There is also still very few of those with 4 stars. So the system is not gonna be conquered that fast, not even by a very few crunchers.

I'm searching how I could add new badges.
A solution could be to increase badges size, so I could add more stars.
For example
:
20 apps || 40 apps || 60 apps || 80 apps || 100 apps | 120 apps | 140 apps | 160 apps | 180 apps



I could also increase credits amounts like this:

100 hours
250 hours
500 hours
1,000 hours
2,500 hours
5,000 hours
10,000 hours
25,000 hours
50,000 hours
100,000 hours
250,000 hours

toms83
 
Send message
Joined: 30 Mar 10
Posts: 4
Credit: 568,247
RAC: 0
Total hours: 1,575,893
Message 1516 - Posted: 11 Aug 2013, 20:45:37 UTC - in response to Message 1345.

I'm searching how I could add new badges.
[...]


So, some kind of decision has been made? Or you still thinking about something different?

Profile Steve Hawker*
       
Send message
Joined: 31 Oct 12
Posts: 125
Credit: 655,634
RAC: 0
Total hours: 1,576,278
Message 1569 - Posted: 4 Sep 2013, 19:33:48 UTC - in response to Message 1449.

I didn't know I was Bragging, I can't seem to Post anything in any Forum without someone suggesting I'm either Bragging or Complaining ... Bye


They're almost certainly British then...

As far as I'm concerned you can post every day to every forum singing your own praises. Takes effort to do what you've done. I'm running six pathetic machines and I could sit there all day just tweaking away. No idea how much time you must spend.

Top effort, mate!!

ps. I'm English so I know of what I speak.

Profile Steve Hawker*
       
Send message
Joined: 31 Oct 12
Posts: 125
Credit: 655,634
RAC: 0
Total hours: 1,576,278
Message 1570 - Posted: 4 Sep 2013, 19:45:41 UTC

As we still dont have a decision, here's my idea based on all the bits of all the ideas that I like.

~ Extend the range of hours (I like the days, years thing personally. I particularly like the WCG system: easy to get to silver/gold, hard to reach ruby)
~ Extend the number of apps in sets of 20
~ Scrap the circle of stars badge design. Replace it with an image for each set of apps, with the background color representing hours (/days/years)

So if we have 10x20 apps with eight color levels, then we can be awarded 10 badges, one for each set, with each badge having a color background. For example, I would currently get five badges with one ruby, one gold, one silver and one bronze.

First badge is novice level, 10th badge is Einstein.

Think WCG style designs. Perhaps we can have images based on scientific illuminati like Newton, Einstein, Hawking and Armstrong. Not really fussed.

This way no-one loses anything. some people stand to gain, everyone has more to aim for, and the system wont run out until we all have quantum computers in our wrist watches.

Profile Steve Hawker*
       
Send message
Joined: 31 Oct 12
Posts: 125
Credit: 655,634
RAC: 0
Total hours: 1,576,278
Message 1571 - Posted: 4 Sep 2013, 19:54:51 UTC - in response to Message 1511.

KPX said:

Edit: I looked at the stats of the largest teams. People, who now have 5 stars, count in single digits.


I just looked at the top 400 by Credits and there were 18 with five stars. I have five stars and i'm not even in the top 1,000.

Profile Conan
       
Avatar
Send message
Joined: 28 Mar 10
Posts: 565
Credit: 1,173,815
RAC: 186
Total hours: 3,242,096
Message 1572 - Posted: 4 Sep 2013, 22:24:23 UTC

Well that total is going up all the time.

There are currently 20 people with all 5 stars.

There are 37 people with 4 stars.

This is in the top 500 people (on RAC).

An enhanced system to what we have is to be considered.

Some people have already gotten enough hours up and applications to have a number of more stars.

Even I have reached 5 stars but turning them all to Sapphire is going to be a challenge.

Conan
____________

Profile skgiven
       
Avatar
Send message
Joined: 7 Sep 10
Posts: 453
Credit: 945,109
RAC: 0
Total hours: 2,101,570
Message 1573 - Posted: 5 Sep 2013, 17:12:53 UTC - in response to Message 1572.

Also, there are 33 people in the top 200 who have at least one Blue star. These people could be crunching towards higher levels of star if they existed. Some people already have more than 10,000h for 20apps. I've even accumulated 12,000h for my top app.

In addition 27 people have a green star as their top level from the top 200 WUProp crunchers. This demonstrates that within the next 6months the number of people with >5000h for 20apps will almost double.

I say this because I will not be chasing a 4th or 5th star. Well, not this year, and maybe not next year either.
However, I would like to reach Blue before the end of the year, and would then like to be able to get some form of recognition for attaching so many systems to WUProp while running by favorite projects.
____________
=sticky tape=

zombie67 [MM]
             
Avatar
Send message
Joined: 30 Mar 10
Posts: 219
Credit: 8,806,089
RAC: 1,418
Total hours: 30,511,517
Message 1574 - Posted: 6 Sep 2013, 4:21:45 UTC

Do we really know how many people have which stars? Or is it only the subset of folks who have turned on exporting that info? How are these numbers being gathered?
____________
Reno, NV
Team: SETI.USA

Profile STE\/E
           
Avatar
Send message
Joined: 28 Mar 10
Posts: 642
Credit: 3,866,385
RAC: 462
Total hours: 20,095,471
Message 1575 - Posted: 6 Sep 2013, 4:33:07 UTC
Last modified: 6 Sep 2013, 4:34:14 UTC

The Top participants Statistics show 33 people in the top 200 who have at least one Blue star, so going by that alone skgiven is correct. But there's no way of knowing what somebody has attained beyond that that has exporting turned off ...
____________
https://signature.statseb.fr/sig-1323.png
https://stats.free-dc.org/badgesbanner.php?cpid=13a87c3a303bcdca4ba0ed600daebb6b

Sellyme
   
Send message
Joined: 21 Jul 13
Posts: 69
Credit: 691,597
RAC: 0
Total hours: 717,632
Message 1576 - Posted: 6 Sep 2013, 11:22:39 UTC

Yeah, I kind of dislike the concept of exporting not being on by default: there's really no good reason for it, and it makes it difficult to make informed discussion on this kind of thing.

That being said, here is my working out:

To get a blue star, you need to return 100,000 hours of work (20 projects; 5,000 hours). To get all 5 stars you need just 10,000 hours of work (100 projects, 100 hours of work).

So anyone who has maxed out either "stat" has at least 10,000 hours of work recorded. A three hour WUProp@Home task returns exactly 7 points. This presents a problem, because a machine running 64 simultaneous tasks (like a quad-Opteron 6272 rig a friend of mine is running) is valued as the same points per hour of running WUProp as my Nexus 4, which only runs 5 tasks at once. That being said, if we assume that the average number of tasks WUProp is monitoring per computer at any one time is 4 (which, looking through Free-DC host stats, is a bit generous: whilst you can run non-CPU-intensive tasks to increase that, unless you have 20 cores, those non-intensive tasks will simply be adding on junk stats before the necessary 19 other projects hit the requisite number of hours), that means that every 7 points you have, you've returned 12 hours of run time. 10,000 hours (the minimum to max out one of the two badge metrics) is thus 833 WUProp WUs returned, and 5,833 points.

That being said, you'd need 8,333 WUProp WUs returned and 58,333 points to have a sapphire star.

I've gone through every single account with in excess of 40,000 points, just to be safe, and here is the # of people with Sapphire Stars:

5 STARS: 2
4 STARS: 3
3 STARS: 3
2 STARS: 6
1 STAR: 33

(Those are inclusive; i.e., STE\/E and zombie67 count in all five categories. The figure is the number of people with at least that many stars. Strangely, no-one has 3 sapphire stars.)

Additionally, going all the way to 25,000 points, 22 users have five stars.

This means that there are 55 instances of someone maxing out a metric (although some of these instances may be the same people). I think it's safe to say that with 55 people maxing out the badges in some way, it's time for an upgrade.

Previous · 1 · 2 · 3 · 4 · 5 · Next
Post to thread

Message boards : Number crunching : More Stars ???


Home | My Account | Message Boards | Results


Copyright © 2024 Sebastien