Hi kctipton

I\'ve saved a link to your BoincStats page for CPDN. We need to keep an eye on what happens to your credits over the next few days.

---

Cpdn news

I lost almost 51,000 points. From 124,993 to 74,006. Though now it says 77,540. There is no way you can call that a small credit loss. Please tell me those points are not lost. They have been earned. Please reply promptly to this because I\'m taking your answer back to the club. There are a lot of clubs that lost over 250,000 points. Seti.USA lost 340,000 points, LAF lost over 1.5 mil ( I\'m actually ok with that ). These numbers are from boincstats. The teams are going to want those points back. I\'m going to want those points back. Patrick

the only thing I can think of is somehow when I did the live archive dump the other day, a few old results got lost, causing some of these 1-2% losses. but I needed to bump up the cpdn credits since as Thyme Lawn pointed out we\'re actually \"under-crediting\" when compared to other boinc projects. So a 5% boost on that seemed about right. So you may have noticed a drop of 1-2% (when the database archive happened) followed by a 5% increase. CPDN credits have never been an exact science (sort of like climate modelling ;-)....

I have to say this... Once earned, the points ARE an exact science. Whatever their amount they should be permanently protected after they have been earned. This is not the only project that has had point issues in the last couple months. MW dumped points, Drug discovery cannot rebuild their point database, though they are still working on it. There was a 4th one that I can\'t remember. If the points cannot be protected then don\'t donate your time. The idea of Boinc has evolved into teams and personal goals, actual hobbies for 10\'s of thousands devoting machines for nothing but crunching. We do not ask for anything in return so those points better be the next most important thing next to the science. The people donating their machines and time have more than earned it. Patrick

[Patrick wrote:]... Drug discovery cannot rebuild their point database ...

... which may be a consequence of \'protecting\' credits.

Projects vary and CPDN has some special requirements: the work units are long compared with other projects and the \'trickle\' mechanism was invented in response to that, both to report some information as the model runs, but also to allow credit allocation throughout the model\'s run and to give something to a model that runs but doesn\'t finish. The credits are therefore calculated by counting trickles, which is computationally burdensome but does mean that talk about credits being \'lost\' is not usually right, unless the trickles have somehow got lost - credits sometimes vary if the database is reworked or something goes wrong, but eventually the calculation settles down again.

In this case, the rate at which credits are awarded per trickle has gone up to better match that given by other projects, so \'lost\' is again the wrong word. If someone loses credits in the long run then that\'s an error, not what the project intended.

I see credits as having some practical uses:

- RAC variations can signal a problem with a particular computer

- RAC differences signal performance issues between computers

- credit totals can be useful in roughly assessing the experience of people who post, allowing a more appropriate response.

When there\'s a credit problem the advice is usually the same: wait.

Hi all,

I didn\'t have as much credits as a lot of people here, but the credits I DID have, dropped from about 16800 to about 12400 over the last couple of days. That\'s a 25% loss, and I don\'t like that. Since then, my credits went up to just over 12500.

What should I do, just sit back and wait for something to happen? Or should someone at CPDN who knows how this works, look at my specifice situation? I mean: looking at all the posts about this subject, it is starting to look like a common problem, and I hope the people at CPDN will come with a common solution for everyone.

Besides, if this IS a common problem, or at least something that appareantly happens a lot, it would be nice if it were mentioned on the \'News\' page.

Regards,
Pete (from Holland)

Patrick , I\'ve seen your CPDN page at BoincStats. We\'ll be watching what happens over the next few days.

PvB
Paul

---

Cpdn news

PvB, please subscribe to the forum News thread which is at the top of the Number Crunching section. If you enable emails from the project in your CPDN account you\'ll receive an email whenever there\'s a new announcement.

---

Cpdn news

It seems that credits ARE a very touchy subject :)
Regards
Masud.

It is. People at SETI are cramming nVIDIA boards into their rigs just to gain more credits. And people at GPUgrid and MilkyWay are doing the same with ATI boards. It is only at AQUA that CPUs are still gaining more credits than GPUs.
Tullio

---

Has anyone whose credit total has dropped noticed any change in the list of computers attached to the project? I\'m not sure yet if that\'s got anything to do with these problems but it\'s something I\'ll take a look at.

Has anyone whose credit total has dropped noticed any change in the list of computers attached to the project? I\'m not sure yet if that\'s got anything to do with these problems but it\'s something I\'ll take a look at.

my climate change has this

GenuineIntel
Intel(R) Core(TM)2 CPU 6400 @ 2.13GHz [x86 Family 6 Model 15 Stepping 6]

my milky way shows this

GenuineIntel
Intel(R) Core(TM)2 CPU 6400 @ 2.13GHz [x86 Family 6 Model 15 Stepping 6]
(2 processors)

so the only difference between the 2 is the last part ie (2 processors)

I lost 3490 credits yesterday and gained 659 credits today

paul

The change that Milo is talking about, is the disappearance of a computer that used to be on your list.

Also, has anyone with a big credit loss merged computers in the last few months?

Has anyone whose credit total has dropped noticed any change in the list of computers attached to the project? I\'m not sure yet if that\'s got anything to do with these problems but it\'s something I\'ll take a look at.

Milo,

As reported in 37695 I originally lost 7,136 credits on a total of 323,223 (a delta of -2.21%), prior to the +5% across-the-board adjustment. It may or may not be a coincidence that the downward revision was returned to my client at the time I reported completion of a HADAM3P result.

As to your question above, both my \"show all computers\" and \"only computers active in past 30 days\" appear to be accurate and unchanged from before the server upgrade. Nor have I merged any machines in the past few months. I actually have one \"old\" listing that could be merged with a current machine, but I\'ve seen problems with that function on some other projects in the past.

---

Best wishes:)

Patrick , I\'ve seen your CPDN page at BoincStats. We\'ll be watching what happens over the next few days.

PvB
Paul

Thanks very much, I\'ll pass that on. Patrick

[Patrick wrote:]... Drug discovery cannot rebuild their point database ...

... which may be a consequence of \'protecting\' credits.

Projects vary and CPDN has some special requirements: the work units are long compared with other projects and the \'trickle\' mechanism was invented in response to that, both to report some information as the model runs, but also to allow credit allocation throughout the model\'s run and to give something to a model that runs but doesn\'t finish. The credits are therefore calculated by counting trickles, which is computationally burdensome but does mean that talk about credits being \'lost\' is not usually right, unless the trickles have somehow got lost - credits sometimes vary if the database is reworked or something goes wrong, but eventually the calculation settles down again.

In this case, the rate at which credits are awarded per trickle has gone up to better match that given by other projects, so \'lost\' is again the wrong word. If someone loses credits in the long run then that\'s an error, not what the project intended.

I see credits as having some practical uses:

- RAC variations can signal a problem with a particular computer

- RAC differences signal performance issues between computers

- credit totals can be useful in roughly assessing the experience of people who post, allowing a more appropriate response.

When there\'s a credit problem the advice is usually the same: wait.

The information is appreciated, Thanks Patrick

Has anyone whose credit total has dropped noticed any change in the list of computers attached to the project? I\'m not sure yet if that\'s got anything to do with these problems but it\'s something I\'ll take a look at.

I had 11896 credits and now I have 2082.
A meagre drop of 82.5 percent.

Easy come easy go.
I can crunch some more to get my 10K target back.

@Milo
Looking at the tasks completed per host via the host list I now see that I am missing a completed task altogether.
I know that I had completed at least 2 to get over the 10K target.
Now the host list only shows 1.

Perhaps that helps to shed some light on this odd result.

Can I also suggest that if you have to do this again some time, please test it first.
Thx.

please test it first

This is like a bullet - to test it, you fire it.

As for your missing task, that\'s the new normal for those that were completed over a year ago. They\'re archived so that they don\'t have to constantly be recounted for the credits each day.

As for your missing task, that\'s the new normal for those that were completed over a year ago. They\'re archived so that they don\'t have to constantly be recounted for the credits each day.

So the 82.5% devaluation was not likely due to lost data then.
Thank you.
That bit was useful.

Thanks for the replies - I\'ll investigate further next week.

Can I also suggest that if you have to do this again some time, please test it first.
Thx.

The week before the server upgrade (from 2-core 4GB RAM to 8-core 24GB with improved RAID array) we performed a test run of the upgrade procedure by performing a full dump of the active database and restoring it on the new server. Then, Carl\'s archive scripts and modified credit calculation scripts were tested on this snapshot. Once we knew these things worked, we shut down the old database and did the upgrade for real.

However, as Les said:

This is like a bullet - to test it, you fire it.

One thing we could not test would be what happens once clients were allowed to connect once more and start sending in users\' data.

To clarify, what has been archived is the trickle information for old models. There were two tables in the database - trickle and result - that were of such huge size that they were too slow to read effectively. The procedure for calculating credits would read these tables each day and recalculate the credits and the change was to keep the old trickles and sum of old credits in the archive and incorporate these into the daily calculation.

According to BOINCstats I lost 1,980 credits on Thursday. That is _exactly_ the amount this task was credited: (from job_log)

1237196323.462999 ue 300923.264536 ct 207917.900000 fe 900000000000000.000000 nm hadam3p_m0nq_1973_2_006033048_2

It was returned on 16 March and was the latest of three tasks crunched on the Quad. The other ones were HADAM\'s completed in January/February.
From Firefox\'s history, I think this was the Workunit page.

---