[Shaktai]

Predictor@Home currently appears to be the most stable BOINC project with more update and better project team responsiveness then SETI-BOINC. This project is in public alpha status, but has already turned in "results" posted on their homepage. According to the project team, while there is a chance for lost credits or even a credits reset, "if the project science required it", currently the plan is for all credits to remain permenent if possible. The project is kind of a cross between dFold and Folding@Home. Current targets are for CASP 6.
Learn more at: http://predictor.scripps.edu.

There are currently 182 teams and 2308 participants.
Team MacNN is ranked #16 in recent average credit and total credit and still rising. Team stats of this moment are as follows:

Name - Total credit - Recent average credit - Country
1) Shaktai - 6712.66 - 345.68 - United States
2) mikkyo - 2132.95 - 189.99 - United States
3) Turias - 1988.97 - 98.08 - United States
4) drweaser - 573.44 - 69.36 - United States
5) pappy - 329.52 - 65.77 - United States
6) sjmielh - 36.92 - 5.82 - Belgium

If you have ever wanted to get in on the ground floor of a project, this looks like a real good one. NOTE: Due to the way that BOINC credit system works, it takes about 2-3 weeks for Recent Average Credit to stabilize, and can take 1-3 days to see your first credit. It is not a project for folks that want "instant gratification" in their stats.

Thanks to Mikkyo, we have a compiled version that will run under OS 10.2.x, Jaguar (The default Mac download only runs under 10.3.x (Panther). As soon as Scott gets home on Tuesday to "kick the server" the Jaguar version will be available for download at http://teamstats.macnn.com/~distcomp..._ppc10.2.8.gz.

Scott, what is the chance of creating a "promo" page for the download, for team recruiting purposes. There is a demand for the "Jaguar client" in both SETI and Predictor, and it might be a chance to do a little recruiting, while providing a service. I could possibly host a mirror at my .Mac homepage as well (need to look into that).

[reader50]

I've reposted mikkyo's compiles for download on a special page, and added the link to the Team Resources thread.

[Scotttheking]

Originally posted by Shaktai:
Scott, what is the chance of creating a "promo" page for the download, for team recruiting purposes. There is a demand for the "Jaguar client" in both SETI and Predictor, and it might be a chance to do a little recruiting, while providing a service. I could possibly host a mirror at my .Mac homepage as well (need to look into that).

If I'm doing it, slim to none. If someone else can do it, probably pretty good. I'm just swamped (well, nothing to do now, but that's because I'm stuck in California and didn't think to bring stuff to do).

[Shaktai]

Originally posted by Scotttheking:
If I'm doing it, slim to none. If someone else can do it, probably pretty good. I'm just swamped (well, nothing to do now, but that's because I'm stuck in California and didn't think to bring stuff to do).

I'll see what I can do then.

[Shaktai]

THE LATEST PREDICTOR@HOME TEAM STATS

Well mikkyo is soaring through the team ranks and I don't know how long I can hold him off, despite my head start. drweaser is also moving along at a good clip. Turias, pappy and smielh are all moving along at a steady pace.

We have added a couple of new team members, derekm and bborofka who should be showing some numbers soon. Upward progress is slow but steady out of 205 teams now.

Name -- Total credit --Recent average credit
1) Shaktai -- 8457.01 -- 445.20
2) mikkyo -- 3460.14 -- 281.03
3) drweaser -- 1385.06 -- 136.16
4) Turias -- 2171.01 -- 98.00
5) pappy -- 603.54 -- 77.13
6) sjmielh -- 81.27 -- 6.91
7) derekm -- 0.00 -- 0.00
8) bborofka -- 0.00 -- 0.00

Many thanks to mikkyo for the Jaguar and optimized Panther clients, and to Reader50 for posting them. Of course it probably means that my position at the "top" will be short lived. Oh well, it's all for the good of the team and what we hope will be "great science".

[bborofka]

Hello,

I just switched over to Predictor from SETI/BOINC. SETI/BOINC is just so messed up right now, I decided to switch to something that seems to have it together a little more but that's still new so I'm not so far behind everyone else. Hopefully my 12" PB and P4 at work will start churning some results soon.

[mikkyo]

The nice part about Seti being boinc, is that you can set up Predictor to be 99% and Seti to be 1%, which should work nicely with seti's issues.

I'm just playing with boinc right now, not sure I'll keep running it, though it is pretty easy to move up through the ranks it would seem.
I'm not sure I like the credit system, but the more of you that crunch with the faster clients the better, as the team benefits.

[Shaktai]

Originally posted by mikkyo:
I'm not sure I like the credit system, but the more of you that crunch with the faster clients the better, as the team benefits.

The credit system lacks "instant gratification" but rewards consistency and persistence, and tends to grow on you over time. It is very cheat resistent. The validation system also improves the overall quality and reliability of the science.

[mikkyo]

I think we could take the team into the Top Ten if more folks jumped aboard.
Those of you running seti (or trying to) should set up predictor too as a backup for when the seti servers are down and/or no work is available.
Macs seem to do pretty well with boinc clients.

[bborofka]

Originally posted by mikkyo:
I think we could take the team into the Top Ten if more folks jumped aboard.
Those of you running seti (or trying to) should set up predictor too as a backup for when the seti servers are down and/or no work is available.
Macs seem to do pretty well with boinc clients.

It seems like, maybe it's just me, but not many people know about Predictor. Is there a cool sig. badge we could make that shows our credits or something? Help spread the word.

[reader50]

I'll look into a Predictor sig tonight, and submissions are welcome.

Is there a consensus that Team MacNN is in Predictor, not just experimenting? If so, drop into the other thread and vote.

For our BOINC experts, has Berkeley produced any new graphics, or is everyone satisfied with the current SETI sig graphics?

[Shaktai]

Originally posted by reader50:
Is there a consensus that Team MacNN is in Predictor, not just experimenting? If so, drop into the other thread and vote.

For our BOINC experts, has Berkeley produced any new graphics, or is everyone satisfied with the current SETI sig graphics?

I'm in predictor 100%. I really like the project. (No I am not going to beta test climate prediction).

Haven't seen any new SETI graphics that I like.

[reader50]

Large Version, for editing or for sigs on other forums:


Small Version, for sigs everywhere:


They have been added to our Team Sigs page, which also gives instructions for how to set a sig image.

For anyone interested in doing Predictor@home sigs, the "Predictor" word is done in the font Copperplate. I didn't determine the fonts used for "@" and "home", but Times and Geneva respectively give close approximations.

[mikkyo]

I like it.
Works for me.

[Shaktai]

The sig works for me too.

UPDATE: We have moved into the #15 team spot (out of 231) in recent average credit and overall. We've been kind of moving back and forth with other teams the past couple of days, but I think we have almost secured it.

wheeles just joined predictor to test the waters. (Yeah!) There are now 2856 participants in the project and 9 on Team MacNN. A drop in the bucket compared to SETI, but at least we have work. The more that join, the sooner everyone's pending credit gets validated.

Just replaced the mobo on my Athlon 2500 box last night. That should help out some.

We're making headway.

[mikkyo]

I expect a lot more people to join our team, especially with seti being so lacking.
I bet by monday we can make a big leap in the stats.

[reader50]

Sig image aliasing problems fixed, also the text on the small version fixed.

[Shaktai]

We are now solidly in the #15 spot overall, and the #14 spot in daily average credit. Still climbing the ranks, and that is with only 9 team members, 8 of which have been granted credit. (wheeles where are you? Hope you are crunching.)

There are now 251 teams with 3115 participants, 2005 of which have been granted some type of credit. Several hundred are new within the last 24 hours.

This project is catching on. Don't know how many computers, but my most recent addition of 2 days ago has node# 7742, and based upon the rate of new members in the past 48 hours it is safe to estimate more then 8000 computers currently. Too bad that isn't 8000 G5 Macs...

[Shaktai]

Still #15 overall, but closing quickly on teams i.motion and Team Ninja.
In Daily average credit we are firmly established at #13
With our current production, we might be able to make the top 10 in a few weeks (barring any major power-ups by other teams) However, we need more help to advance beyond that. Still at 9 registered team members, 8 of which are showing currently active.

There are now 3250 participants registered and 258 teams.

Predictor@Home, like Distributed folding, is involved with CASP 6. In a sense then, the predictor team and dFold team are in competition. It will be interesting to see which process delivers the most interesting results in CASP. Each project takes a different approach, to the same goal.

While Distributed Folding works on one target at a time, and must change targets every week (requiring weekly user updates) Predictor works on multiple targets at one time and requires no user update unless there is an improved core client released (one is sheduled for the near future) and even then results from older core clients will still be usable.

If you are one of the current dFolders who is not able to keep up with the weekly updates, please consider joining the Predictor team for the duration of the CASP 6 competition which will run until early September.

Both projects are worthy of our support, and the CASP 6 competion period is important to both. I encourage participation in one or the other until September. It is a great chance to be out on the leading edge of science, and to help develop processes that will benefit not only our own generations, but our children's as well. And of course to have some fun doing it.

So let's get crunching.

[Shaktai]

In the last hour, Team MacNN passed Team Ninja and yesterday passed i.Motion to take the #13 spot overall, and is #13 in recent average credit. Since recent average credit is still increasing (it takes about 3-4 weeks to stabilize) we should be able to reach #11 or 12 without problem. With just a little more help we could make the top 10 out of 268 teams. (3435 registered participants, and increase of 185 in the last 24 hours)

We still have only 8 members earning credit. There must be be some more power out there somewhere. Surely someone would love to have the chance to knock me out of a #1 spot again. I've got this big huge target on my back, but no one is taking aim.

[DrBoar]

I am thinking about tossing my Macs into Predictor so...
How do you do dual CPUs?
Is it two folders and then the " -local" argument as in F@H CLI??
TIA

[mikkyo]

nope.
You just run one client and it is smart and runs 2 workers.

[Shaktai]

For those who are interested, Predictor@Home has posted their the top 5 candidates of their first 5 proteins for the CASP 6 competition.
http://predictor.scripps.edu/gallery.php

3 others are due to be submitted today or in the next couple of days. They will likely present the top 5 candidates from each of those as well, in the same gallery. The last 24 hours saw the last minute release of CHARM units. With predictor, the intial targets are ran once, then if time allows, top candidates may be released again for "refinement" as charm units. The project team seems excited about the results they are getting.

As of today Predictor has 281 teams with 3775 participants.
Team MacNN is #13 and climbing slowly. Yours truly is #17 out of all participants.
Predictor@Home Team ranks

Oh, and SETI-BOINC is down again tonight.

[Shaktai]

Originally posted by DrBoar:
I am thinking about tossing my Macs into Predictor so...
How do you do dual CPUs?
Is it two folders and then the " -local" argument as in F@H CLI??
TIA

In the "website" preferences, you specify the number of CPU's to use. The default is up to 2, but you can specify more or less. One participant is running a "data center server" with 32 -650mhz Pentiums. Suffice to say that is currently the fastest computer on the project. The #2 computer is a quad 2.8 ghz Xeon.

[mikkyo]

Team MacNN is now in 12th, soon to be 10th place.

[Shaktai]

Originally posted by mikkyo:
Team MacNN is now in 12th, soon to be 10th place.

Now we are #11 in recent average credit, and just briefly slipped back to #14 overall. However with our current production, mikkyo is right.. #10 is within striking position. Just a matter of days.

For all MacNN participants, make sure your are running mikkyo's re-compiled versions. They are faster. (but totally legal) Hopefully 3.20 versions will be up soon.

Smaller, faster & better. mikkyo re-compiled versions. Thanks again mikkyo.

And welcome to new team members Steve Whitcliffe, Jason Harris and Dr Boar. You guys are making a big difference already.

[mikkyo]

That is daily production if that wasn't clear.
Soon the Totals will match as well.

11th now in Daily.

[bborofka]

[Shaktai]

We are #10 in recent average credit and still climbing slowly, but have risen to #13 overall.

There are 338 teams now, with 300 having returned credit. 4659 total users.

It looks like by tomorrow morning I'll be choking on mikkyo's dust as he roars past me in recent average credit, and not that far behind in total credit. I knew it was only a matter of time, but thought I might hold him off for a few more days.

Now where are those 6 dual G5's I was keeping in reserve. Oh yeah, only in my dreams.

It is all good for the team though.

[Shaktai]

(cough, cough, gasp) Excuse me the dust hasn't quite settled since yesterday. mikkyo roared past me yesterday in recent average credit with the souped up, hot rod, farm of his. (cough) Before too many more days, I expect him to pass me in total credit. (choke, cough - damn dust).

Any way (cough), we are still #10 in recent average credit, and #12 overall. The going gets tougher from here on out, but we can still make some headway. We would need to double our current power in order to take the #6 spot. Fortunately, most of the team is still seeing gradual increases in recent average credit, so we haven't reached our peak yet.

For the new-comers, it will typically take 3-4 weeks for your recent average credit to "max out", so don't get impatient.

To clarify, Recent Average Credit represents our current Average production rate, as opposed to total credit.

As of this posting. 357 teams, with 4948 participants and a whole lot of computers. Most of them are turning in work. mikkyo is ranked #15 out of all users in DAC and #30 in total credit. The Predictor project folks would like to see more Mac Users, to help clear up some of the pending credit.

[mikkyo]

The more folks that run Predictor on the mac, the faster we get credit, so team players really help out, and mac users really help mac users.

[DrBoar]

FYI
The 10.2.8 client actually runs as low as Darwin 1.3.7 AKA 10.0.3.
So with a 200 MHz 604E and a 256 L2 cache and 184 I need several of those to catch up with mikkyo

[The Raven]

Has anyone run BOINC as a service, is it built in the client or do I need firedaemon.

[Shaktai]

If you have machines running Charm work units, check them and make sure they aren't stuck. There are problems with charmmB120_3.10 application on Macs. You may experience "stuck" work units or errors. Charm units should go quickly, and I had one that processed for over 18 hours and still didn't finish. It just errored on a restart.

STOP BOINC
Restart using the -update_prefs http://predictor.scripps.edu

That should download the new charmmB120_3.11 for Mac that should fix the problems. (Just released a little bit ago).


IN OTHER NEWS
By some miracle beyond my understanding, I have jumped ahead of mikkyo again. Haven't a clue how. Maybe some of his farm got stuck on those "charm" units that won't process.

We remain at #10 in recent average credit, but have risen to #11 in total credit. We should reach #10 in total credit before long. However, we need at least 30% more power to go beyond #10.

Einstein@Home has released its "beta project" on BOINC, and it is called Pirates@Home. Just one more project moving to BOINC. Pirates is "very beta" though, so don't expect any stability out of it yet.

[mikkyo]

Predictor folks should let me at their worker source code...
I'm working on the silly seti worker to see if I can get a better optimized one running, until then.
A few of my machines were busy with other stuff the last few days.
I doubt I'll actually catch up to Shaktai any time soon.
Well maybe if I put the Tiger preview on all my machines and run the optimized version for that...

[Shaktai]

Originally posted by mikkyo:
Predictor folks should let me at their worker source code...
I'm working on the silly seti worker to see if I can get a better optimized one running, until then.
A few of my machines were busy with other stuff the last few days.
I doubt I'll actually catch up to Shaktai any time soon.
Well maybe if I put the Tiger preview on all my machines and run the optimized version for that...

Don't think they'll open up the Predictor source code any time soon, and at least not until after the CASP competition. Too many bigger concerns at the moment.

Speaking of, the Precictor forums and stats are down, and will be until tomorrow, but servers are finally back up. Work units are limited though. No new ones until they test the new modifications Mac & Linux machines may stand a better chance of getting work then windows machines since there are plenty of Mac units needing validation.
-- Here's your chance to pass me again mikkyo All my windows boxes are "out of work".

Can't get to the SETI site either. Don't know if they are down "again" or "still".
Signed up for Pirate@home as a very low priority, but they don't have any work either. Guess my poor computers will get some rest, except for the iBook which still has plenty of work.

[mikkyo]

I got the Seti worker building again, optimized workers coming tomorrow most likely for testing. If they work well we will have yet another Team MacNN offering for the masses.

[sjmielh]

Will you make your source code available online?

Sjmielh

[mikkyo]

The clients are all just compile-time cpu-specific optimized builds that anyone can make with the public project source code.

You can pass gcc flags to configure when you initially configure the boinc/seti clients.
Flags like -O3 -faltivec -mcpu=G5 -mtune=G5 -funroll-loops -ffast-math -fstrict-aliasing -fastf -arch ppc970 will yield a heavily optimized version of the client only for a G5 based machine.

If you have OS X DevTools installed on your machine, see
file:///Developer/Documentation/ReleaseNotes/DeveloperTools/GCC3.html
for more info.

[mikkyo]

Team MacNN is solidly in 10th place now in both average and total credit.
And we are still climbing in output since it takes weeks to show up.
Many new folks have joined and the prospects look good.
In a few days we will know how much better things can get.
Let's keep the crunch numbers up and add some more folks and go for the top 5!

[Shaktai]

Originally posted by mikkyo:
Team MacNN is solidly in 10th place now in both average and total credit.
And we are still climbing in output since it takes weeks to show up.
Many new folks have joined and the prospects look good.
In a few days we will know how much better things can get.
Let's keep the crunch numbers up and add some more folks and go for the top 5!

BTW, right now work units are scarce for Windows machines at the moment, so it is a good time for a Mac push. Lots of mac units out there needing validation.

My iBook is crunching steady, but only one of my Windows boxes has any work at all.
-- Predictor - very little work for Windows
-- SETI - Only very sporadic work if any, but should change soon
-- Pirate@home. - with Predictor & SETI down, they got slammed beyond what they were prepared for. Basically down for the time being.

Predictor now has 6410 participants with 398 teams. They have temporarily shut down account creation until they get caught up.

Oh, and without my Windows boxes crunching, mikkyo quickly dashed ahead of me and now owns the #1 team slot.

[Shaktai]

Well, after a hiatus to fix back-end issues, Predictor is back up and running. There is still a glitch with the validation of units however, because the validator is apparently not working correctly.

QUOTE: Michela Taufer, one of the predictor developers
The transitioner and validator processes have been down for a while as well and have quite a queue of results to process. That's why it might take a while before results are validated and credit granted. Eventually it will be done, but it might take some time.

Thanks

Michela

So be patient. Recent average credit appears to be dropping for everyone, but it is just a temporary issue. All teams/individual are being impacted equally.

[finizio]

So be patient. Recent average credit appears to be dropping for everyone, but it is just a temporary issue. All teams/individual are being impacted equally.

I have a problem with the two computers running Jaguar.
With your client the completed work units claim 0 credit and obtain 0 credit.
Anyone has the same problem?

yours
Odontoinformatica

[mikkyo]

Predictor credit can take at a minimum 3 days and sometimes as longs as 3 weeks to show up. You just have to keep crunching and wait for others to complete the same workunits for validation, which then gives you credit.
As long as you aren't getting errors running the client you will be fine.

BTW, you must be running the 3.20 client as the 3.19 client is not valid any more.

[finizio]

Originally posted by mikkyo:
Predictor credit can take at a minimum 3 days and sometimes as longs as 3 weeks to show up. You just have to keep crunching and wait for others to complete the same workunits for validation, which then gives you credit.
As long as you aren't getting errors running the client you will be fine.

BTW, you must be running the 3.20 client as the 3.19 client is not valid any more.

Thank you for your answer.

I'll wait, but my Panther client claims credit and credit from Panther are listed in claimed credit link.
Jaguar client (3.20) reports completion without errors but always 0 claimed credit and the results seems to disappear in the claimed credits page.
Here's a piece of my results page for a Jaguar client.

882266 236538 24 Jul 2004 17:20:06 UTC 26 Jul 2004 16:54:20 UTC Over Success Done 0 0.00 0.00
882168 236506 24 Jul 2004 17:20:34 UTC 27 Jul 2004 0:49:31 UTC Over Success Done 0 0.00 0.00
881862 236417 24 Jul 2004 17:20:29 UTC 27 Jul 2004 0:17:00 UTC Over Success Done 0 0.00 0.00
881186 236193 24 Jul 2004 17:19:54 UTC 26 Jul 2004 16:20:32 UTC Over Success Done 0 0.00 0.00

[Shaktai]

finizio,

What type of work units are those? Are any of them t0235?

Can you list the actual target (work unit) types? Also, if you click on the work unit ID itself, does it show any type of error in the information?

[finizio]

Originally posted by Shaktai:
finizio,

What type of work units are those? Are any of them t0235?

Can you list the actual target (work unit) types? Also, if you click on the work unit ID itself, does it show any type of error in the information?

Here's one summary of a workunit

Summary of workunit

application
MFold Protein Structure Assembly

created
25 Jun 2004 17:34:52 UTC

name
t0203C_1_387

granted credit
0.00

And here's the errors in the result's page

Result ID
907492

Name
t0203C_1_387_6

Workunit
19191

Created
24 Jul 2004 19:01:46 UTC

Sent
25 Jul 2004 12:59:29 UTC

Received
27 Jul 2004 7:47:48 UTC

Server state
Over

Outcome
Success

Client state
Done

Exit status
0 (0x0)

Host ID
16347

Report deadline
8 Aug 2004 10:19:29 UTC

CPU time
0

stderr out
<core_client_version>3.20</core_client_version>
<message>file transfer error: couldn't get input files:
t0203C_b1_387: file was not found on server

</message>
<active_task_state>0</active_task_state>
<signal>0</signal>


Validate state
Initial

claimed credit
0

Granted credit
0

Client version
---


Here's a different error

Result ID
882266

Name
t0233C_1_5154_1

Workunit
236538

Created
23 Jul 2004 7:39:17 UTC

Sent
24 Jul 2004 17:20:06 UTC

Received
26 Jul 2004 16:54:20 UTC

Server state
Over

Outcome
Success

Client state
Done

Exit status
0 (0x0)

Host ID
16340

Report deadline
25 Jul 2004 17:20:06 UTC

CPU time
0

stderr out
<core_client_version>3.20</core_client_version>
<stderr_txt>

</stderr_txt>


Validate state
Invalid

claimed credit
0

Granted credit
0

Client version
---

[finizio]

btw

Jaguar computers tried to process more than 40 units without any claimed or granted credit!!!

[drweaser]

Originally posted by mikkyo:
Predictor credit can take at a minimum 3 days and sometimes as longs as 3 weeks to show up. You just have to keep crunching and wait for others to complete the same workunits for validation, which then gives you credit.
As long as you aren't getting errors running the client you will be fine.

BTW, you must be running the 3.20 client as the 3.19 client is not valid any more.

About how much pending credit do you normally 'sit' on? I am sitting on about 3500 pending credits right now and it seems to be growing a pretty steady ~500+ each day. I am getting a fair amount granted each day, but pending is still outgrowing it pretty wildly....

Of course, I did just convert ~25 eMacs over to Boinc/P@H/S@H over the last couple of weeks, but surely it should even out.......at P@H is giving out a steady stream of work!

drweaser

[Shaktai]

Originally posted by drweaser:
About how much pending credit do you normally 'sit' on? I am sitting on about 3500 pending credits right now and it seems to be growing a pretty steady ~500+ each day. I am getting a fair amount granted each day, but pending is still outgrowing it pretty wildly....

Of course, I did just convert ~25 eMacs over to Boinc/P@H/S@H over the last couple of weeks, but surely it should even out.......at P@H is giving out a steady stream of work!

drweaser

Well, I was up to about 4500, and my recent average credit was dropping still. dropped all the way down to 737. However tonight, pending credit is down to 4200, and my average credit has climbed back up to 758, and appears to be rising again. Hopefully they got the validator issues worked out.

25 eMacs? Looks like I need to keep an eye on my rearview mirror.

UPDATE: Did some quick math. with what you already have plus those eMacs, that should be enough to cleanly propel us into #9 spot and put 8, 7 and 6 within reach. - Nice!!!