[Shaktai]

WOW!!!

The two of you have given the Climate Prediction team a huge surge in prodution and credits overnight. Welcome to the best Mac team ever.

http://climateapps2.oucs.ox.ac.uk/cp....php?teamid=62

polishjeff, that is one very nice Mac farm you have running there. Of course it means I have now been knocked down to #4 with drweaser hot on my tail, but with that kind of team production, I really don't mind.

[Knightrider]

Hi Shaktai,

Thanks for the welcome. I have been looking for a mac home for a couple of months, since I got my G5 with 2.5gb ram. I have projects in mind but nothing started yet. I need to relearn the mac OS, so still a newbie on OSX. (I also run a PC with a 3.06 intel cpu)

I have been trying out the G5 optimized versions. No luck with the
seti_Worker_G5Optimized_altivec (I dropped the files in the right place and it runs, but a wu is taking about 7 hours) so I will try the G5+setiworker from this page:

http://members.dslextreme.com/~reade...eam/boinc.html


I am still a novice with Terminal, but have managed to get both SAH and CPDN running.

Any tips or advice always appreciated.

K.

[Shaktai]

7 hours for SETI is way too long, on a G5. When you added the SETI worker, did you also include the app.info.xml file? That file basically tells the program which SETI worker to use, otherwise it will default to the regular unoptimized SETI worker. I believe that it is required for both versions. Javalizard's altivec enhanced version should be faster.

The terminal gets easier with practice, that much I can say. I am still a terminal novice, but manage to handle the basics without difficulty. We have some very knowledgable folks here though, so there is always someone who can help figure things out.

[Knightrider]

Shakti,

Thanks for the reply.

I have posted a question about the altivec version on this thread:-
http://forums.macnn.com/showthread.p...02#post2193502

I too, think 7 hours is to long - even when runnibg CPDN in a seperate Terminal window.

K.

[Welnic]

The best way to see what is going on with these issues is to use top, which shows you the running processes and how much of the processors each is using. I use the command:

top -u -s 20

The -u sorts by processor use and the -s 20 means that it only updates every 20 seconds. A 5 second update works okay if you are only looking for a short time. I usually just have mine running.

If you are running both seti and cpdn separately then they will both be running two copies because you have two processors. Which means that they will be fighting for processor time. You can watch top to see this. I would guess that they would be more or less equal. You also might be able to tell exactly which seti work unit it is using.

Boinc is set up to share between two projects like this. If you go to the boinc folder that you are running cpdn in, stop that client, and then run it again with the flag -attach_project you can add seti. Then the client will pause seti for an hour while it runs cpdn, and then pause cpdn while it runs seti. This is more efficient than running them both at the same time.

If you do this you need to go to one of the sites and set the line:

Leave applications in memory while preempted?

in your general preferences to yes. This makes the transition between the two easier.

[reader50]

Welnic, a helpful note. You can often combine control flags, top is an app that allows it. ie:

top -u -s 20
top -us 20

... or what I use most often to reduce top's own CPU usage:

top -uds 10

[Knightrider]

>The best way to see what is going on with these issues is to use top, which shows you the running processes and how much of the processors each is using. I use the >command:

>top -u -s 2

I use the 'Activity Monitor' in 10.3.5; but here's what I got: with 'Top" in Terminal-

Processes: 57 total, 6 running, 51 sleeping... 138 threads 08:30:27
Load Avg: 4.33, 4.15, 4.05 CPU usage: 92.8% user, 7.2% sys, 0.0% idle
SharedLibs: num = 114, resident = 33.9M code, 3.41M data, 9.90M LinkEdit
MemRegions: num = 9648, resident = 220M + 9.93M private, 95.0M shared
PhysMem: 217M wired, 132M active, 413M inactive, 763M used, 1.75G free
VM: 4.71G + 79.1M 28407(1) pageins, 9874(0) pageouts

PID COMMAND %CPU TIME #TH #PRTS #MREGS RPRVT RSHRD RSIZE VSIZE
398 Terminal 46.7% 2:53.13 6 89 192 8.48M+ 25.5M+ 30.6M+ 242M+
1308 setiathome 37.4% 1:51:08 1 11 38 13.8M 1.51M 13.6M 42.4M
1309 setiathome 37.1% 1:46:09 1 11 39 14.8M+ 1.51M 15.7M+ 43.4M+
412 hadsm3um_4 36.9% 18:58:35 1 12 36 41.1M 7.55M 45.8M 91.2M
411 hadsm3um_4 36.3% 18:57:32 1 12 36 41.3M 7.55M 46.0M 91.2M

SAH & CPDN are usually up around 50% - i.e. 50% of each cpu.

>Boinc is set up to share between two projects like this

The problem I have with attaching to Two projects in the single Terminal window is that I start one and its up and running. I havn't been able to work out how to attach to another project after that, other than a new Terminal window.

>Leave applications in memory while preempted?

This is set.

What I would like to be able to do, is to run one project on each cpu. As I understand it, if I change my prefs to one cpu and run two projects, then both projects will still only use one cpu i.e. 50% each project/wu. Setting prefs for Two cpu's means that each project will download Two wu's per project, so I still only get 50% for each wu.

Is there a work around for this?

Thanks,

K.

[Welnic]

If you run ./boinc -attach_project it will ask you for the URL of the project that you want to add. Run that in the boinc folder that you already have cpdn running. It will then add seti and then if you just start it up it will run both seti and cpdn. Once you attach a project then it will run that project also whenever it starts up.

One of the advantages of this, other than you just have one client to deal with, is that if there is no work for seti it will just run full time on cpdn untill there are seti workunits again.

To summarize:

1) Copy the boinc client into a folder. I use my home folder for things like this instead of Applications since they are always run from the command line and the terminal starts in your home folder.

2) I rename the client boinc, so that when I write a cron job to start it I don't have to change it if the boinc version changes. If I need to know the version I run ./boinc -version.

3) If the client won't run I do:

ls -alrt (or) ls -l

if boinc looks like this:

-rw-r--r-- 1 clint clint 1137820 10 Sep 23:26 boinc

then I run:

chmod +x boinc

then it should look like this:

-rwxr-xr-x 1 clint clint 1137820 10 Sep 23:26 boinc

and actually work.

4) Then start the client up with ./boinc. Then paste the URL and then the id that you get when you register for a project. Then watch to see that it successfully downloads both some data and the projects worker. If it seems to be going okay then use top to check to see that it is running two workers so that both processors are being fully used.

5) Then stop the client with Control-C or Command-.. After it stops start it again with:

./boinc -attach_project

Then paste in the new URL and id. Check to see if that runs okay. When there are two projects then of the four workers (on a dualie) only two will be running at any one time, with that changing around once an hour.

6) Stop the client again and then start it the way that I plan to using cron:

nohup ./boinc > output.log &

nohup keeps it from being killed if you close the terminal window or log out. This is very necessary on a linux box, but it doesn't seem to matter that much in OSX. I found one situation that it makes a difference and I don't remember what that was. I run it anyway.

> output.log puts the output, which you normally see in the terminal, into the file output.log. If you do >> output.log then it will append it onto the end of file, this is useful if you are having problems as the old file doesn't get overwritten. Otherwise the single > keeps the file from just getting bigger and bigger.

The & means that the process runs in the background, you get the terminal back right away.

You can watch the output.log by running:

tail -f output.log

but I have found that using the console takes up less cpu time and you can scroll up easily to see the whole file. If you just double click on any file with .log it will open the console.

7) Then I run:

crontab -e

and add the line:

@reboot cd ~/boinc/;nohup ./boinc > output.log &

The @reboot means that it runs when ever the machine first starts up, even before you log in. The ~ means my home directory, and then the next part starts the action.

So at this point the client is running on both processors, if one project runs out of work the other one will run automatically. If the machine reboots then the client will start up right away. If I want to look at what is going on then I open console and look at output.log. If you quit console with windows open then they will be opened the next time that you run it.

[Shaktai]

Originally posted by Knightrider:
>The problem I have with attaching to Two projects in the single Terminal window is that I start one and its up and running. I havn't been able to work out how to attach to another project after that, other than a new Terminal window.

Well here was one I could answer, but Welnic beat me to it. Simply speaking though.
1.) Start one project
2.) Hit Command . (period symbol) to quit the processing.
3.) Restart the project with the -attach_project command like Welnic said. The first project is already there, so now add the second. BOINC is just a framework. The actual projects run inside it, so it will run any attached projects.

What I would like to be able to do, is to run one project on each cpu. As I understand it, if I change my prefs to one cpu and run two projects, then both projects will still only use one cpu i.e. 50% each project/wu. Setting prefs for Two cpu's means that each project will download Two wu's per project, so I still only get 50% for each wu.

You control this by setting Resource Share in the preferences for each project. If you set each project at an identical resource share, like 50, then each project will get 50% of the time. Each number is just a ratio to the other projects, so you can set any number you like. For instance 200 for one and 100 for the other would work out to 66%/33%, etc.

BOINC is pretty smart, and after a couple of days once it knows what the computation times are for each project, it will settle right in to the ratio you set. With a dual processor, most of the time it will run one on each CPU. However, don't worry if the same project is on both. BOINC will periodically interrupt one, to run the other, to keep the actual CPU time at whatever ratio you set for multiple projects.

An example is that Predictor is my preferred project. I want it to run almost all the time when there is work. Predictor is set to 300, Climate Prediction is set to 10. If there is no work from Predictor, Climate will run 100% though. However, if is has 3 projects, and I set 300-10-10 then when Predictor has no work, it will share 50/50 with the other two projects.

It is also smart enough to recognize deadlines. If a particular work unit is reaching its deadline, it will give it more time, so the deadline can be met.

Pretty Cool really. Not 100% accurate, but very close with the latest versions of BOINC.

[Capt Shane]

I just joined up hopefully we will be kicking ass soon :-)

[mikkyo]

It always good to see more folks join us.
Welcome to the team.
Wait a minute, why am I in 3rd place??

Looks like everyone got to the questions already

We are such nice helpful folks

Here is my current share: Predictor 10000, CP 2000(for now), Seti 1

After Predictor comes back, CP will goes back to 1000.

I think the scheduling share is a bit funky due to the imminent due date preemption.
I think a Crunch XX til no work, then crunch YY til no work, then ZZ, might be better, with user controlled dropping, like if work for XX, forget about Seti work in progress.

[Knightrider]

Special thanks to all for the help.

Now running cpdn and sah in same Terminal window + learnt some about it.

Resources set to cpdn 300 and sah 100, so I will see how that goes for now.

Thanks again.

K.

[Shaktai]

Originally posted by Capt Shane:
I just joined up hopefully we will be kicking ass soon :-)

Welcome aboard. Every bit helps.

[mikkyo]

Thanks to all the new folks, Team MacNN is now in the top 20.
We should be moving up a few ranks over the next couple of days.

http://climateapps2.oucs.ox.ac.uk/cp.../top_teams.php

[CPU_Angel]

I just joined Team Macnn a few days ago. I have been looking for a team , for a long time. I hope I can help.

[mikkyo]

Originally posted by CPU_Angel:
I just joined Team Macnn a few days ago. I have been looking for a team , for a long time. I hope I can help.

Why, yes, yes you can.

Welcome and we are now 19th with 18th several hours away.
Once again thanks to all the new folks.

[mikkyo]

I most humbly apologize.
My last statement about moving up in several hours was completely incorrect.
I'm very sorry.
We are already in 17th and should see 16th soon.


Forgive my way too conservative estimate of our crunch power.

[Shaktai]

Originally posted by mikkyo:
I most humbly apologize.
My last statement about moving up in several hours was completely incorrect.
I'm very sorry.
We are already in 17th and should see 16th soon.


Forgive my way too conservative estimate of our crunch power.

That's #17 "overall" and #11 and still climbing in RAC Recent Average Credit. Apparently all my computers have been crunching even while I'm gone. Somehow I actually snuck ahead of Mikkyo. I didn't expect that.

Thanks to all the new team members that are "making a difference".

[Welnic]

How does cpdn deal with credit when you change teams? Does the credit that you currently have stay with the old team? Do you still show up in some fashion at the old team? Definitely if you total up the members total on Team MacNN that is much larger than the team total.

Just askin', by the way.

[reader50]

Earlier, SETI@home Classic had the annoying habit that when you changed teams, they did not automatically retotal the team. Your credits moved with you, but it could be months until the next database repair caused the team to be retotaled.

The problem was eventually fixed.

[mikkyo]

I have no idea how Boinc projects handle changing teams.
I would think the credit would stay with the team, and new work would go towards the new team.

We are now in 16th.

[Welnic]

F@H is very straightforward about how the credit with team switching works and that makes it easy to understand. You can see just looking at your stats that you could have points that you did on separate teams. It seems that most projects that let you move them around just don't mention it, like the boinc projects don't. But the totals on Team MacNN don't add up at present.

[Shaktai]

With BOINC, credit you earned while a member of the team, stay with the team, but the team does not get credit you earned prior. We have several members of Team MacNN, including our #1 member, who had crunched units either as individuals for no team, or as members of other teams. Since they joined the team, we now get credit as long as they are with the team, and the result is that members total credit is "much" greater then team credit.

Of course, an individuals total credit always stays with the individual.

[Thyme Lawn]

Your increased production has been noted here

I hope some of you are going to pop by our forums and give us a wave as you go racing by next week

[Shaktai]

Originally posted by Thyme Lawn:
Your increased production has been noted here

I hope some of you are going to pop by our forums and give us a wave as you go racing by next week

Will have to do that.

We are now #15 overall and #10 in recent average credit, which reflects where we can be eventually with our current production. Team Picard, linked above, is currently #13, but we are strongly outproducing them by more then 2200 per day, but trail them by 15,000+. Our RAC is still climbing, so there is a good chance we will pass them in under a week.

[Thyme Lawn]

Originally posted by Shaktai:
Team Picard, linked above, is currently #13, but we are strongly outproducing them by more then 2200 per day, but trail them by 15,000+.

You're just over 7,000 behind us now. We would prepare ourselves to repel your boarding party, but I guess you're not going to hang around long enough for that to have an effect

At the current rate I wouldn't be surprised to see you in the the top 10 before too long

[Shaktai]

Originally posted by Thyme Lawn:
You're just over 7,000 behind us now. We would prepare ourselves to repel your boarding party, but I guess you're not going to hang around long enough for that to have an effect

At the current rate I wouldn't be surprised to see you in the the top 10 before too long

Wow Our production is still going up.

We are only 2000 behind Team Picard now, and should pass them within 24 hours. Our RAC shows us headed for #9, but our actually daily production appears to be much higher then that. As of right now, our total credit is 205443.25. I'll try to check it again in about 24 hours to get a "real" idea of our production. Seems that mikkyo isn't the only one who underestimated our crunch power. The RAC calculation can't keep pace with our real increase.

[Welnic]

That RAC seems to be totally useless to me. One of my computers has been running constantly on cpdn, turns in about 400 points a day and has been doing that for the last 24 days. Its RAC is 400. Another one I had running from Sept. 10-15, then I had to switch everything back to dFold to hold onto 10th place. It turned in 1436 in those 6 days, which is 240 points a day. It hasn't done anything since then and its RAC is 250. I don't know whether the formulae is messed up or it is just miscalculated somehow.

I have been looking at stats at:
http://www.boincstats.com/

Looking there I can see that Team MacNN is sixth in production for the last week. Not bad.

[Thyme Lawn]

Originally posted by Welnic:
That RAC seems to be totally useless to me. One of my computers has been running constantly on cpdn, turns in about 400 points a day and has been doing that for the last 24 days. Its RAC is 400.

RAC is a total waste of space as far as CPDN is concerned. The total credit is the only thing I watch.

I've seen RAC defined as being an exponentially fading average with a halflife of a week, but the only way I'm ever going to understand what that means is by digging into the BOINC code. I don't think I'm that bothered

BTW, only 600 to go now

Updated 3 hours later ...

Congratulations Team MacNN

You're now 600 ahead of us. Congrats thread created on our forums here

[mikkyo]

Too bad we can't assimilate Team Picard
We are a peaceful race though, and passed just like ships in the night.

As for RAC:
You turn in 10 units a day.
Those 10 units get handed out to other people to crunch for verification at some unknown time between before and after.
A week goes by and you have 70 units turned in, only about 12 have been verified and only 4 in the last 12 hours, so you get a RAC like 4.
You continue to stack up units completed, and results continue to randomly get verified by other crunchers.
You turn off the machine.
Now you have, say, 40 units left to verify out of a total of 100, and those credits will trickle in over time, giving you an increasing and decreasing RAC to match the randomness of others completing your units.
At least that seems to be the idea for Boinc.
Each project will have different oddities, but RAC seems to really be a measurement of how many other folks complete some of your units.
So yeah, not really useful, but if the whole system were running perfectly, with no lost crunchers and no dropped units, then RAC would be a somewhat useful and predictable guage of how well the validation system works.
Heh.

ps Several of my machines have been down for a week, yet my RAC hasn't changed much.

[Welnic]

Except that in cpdn they don't have the verification crap, you just get the credit when you trickle. But it still doesn't make any sense.

Not that I'm bitter about the verification, but I have turned in a couple of dozen work units for seti and haven't seen any credit.

[Shaktai]

Well, RAC be damned, my calculation is about 14,600 per day (or actually 20 minutes shy of 24 hours). That looks like with current production, we might eventually take the #5 spot.

Mikkyo, kind of figured some of your machines were either down or on another task. No other way I could have passed you. Now you get those other boxes cooking, and we might even go for #3 or 4. Better yet, let's just get a half dozen more team members.

[Shaktai]

Well Phoenix Rising is next to fall, in another hour or two, followed by boinc-irc before another day is passed. Ars Tech-Baked Alaska and Team AnandTech will both follow in short order, leaving us at #10 overall. It gets a little tougher from there on out, but not hugely show. #5 is within reach.

[Thyme Lawn]

Originally posted by Shaktai:
Well Phoenix Rising is next to fall, in another hour or two, followed by boinc-irc before another day is passed. Ars Tech-Baked Alaska and Team AnandTech will both follow in short order, leaving us at #10 overall. It gets a little tougher from there on out, but not hugely show. #5 is within reach.

TPR are now history

Boinc-irc won't stay ahead of you for long and TeAm AnandTech are likely to be next to fall to the Team MacNN juggernaut (Ars Technica - TBA should stomp them before you catch up, but their celebrations will be short-lived ).

[Welnic]

I'll start cranking on Tuesday once dFold is over. I have to keep going over there to hold on to tenth.

[Shaktai]

Well, looks like about 24 more hours and we'll pass TeAm AnandTech for the #11 spot.

[Shaktai]

Team AnandTech is behind us now, and we are in the #11 spot. Another 24 hours and we should pass Ars Technica - Team Baked Alaska, to make the top 10. The going gets tougher after that, but we should still be able to move a up a few more slots.

I still miss Predictor though....

[mikkyo]

We are now in 10th and still growing.

[Shaktai]

Originally posted by mikkyo:
We are now in 10th and still growing.

Woo Hoo!!

Hey mikkyo, you better crank it up a couple of notches or get out of the way, because it looks like Welnic intends to steamroll right over both of us and drweaser.

It is starting to look like the team might be able to go for #2, and I might venture to say that #1 isn't totally out of the question in the long run.

It's getting hard to keep track of all the new folks, so welcome to everyone.

[mikkyo]

I did crank it up, you see our dump for the day?

http://www.setisynergy.com/stats/teams_cpdn.php

About time Welnic got back into the fray

Now if I could just get cobramac to switch over.

[Scotttheking]

Soo, how do I run this?
I've got OSX 10.2.8, Linux, win98 (NOT MY FAULT), and win2k. Win2k of course needs to be CLI so I can make it a service.

[mikkyo]

10.2.8 wont work as far as I know.
I havent had time to try making a 10.2.8 client.
Other than that, go here, register,
http://climateapps2.oucs.ox.ac.uk/cp...count_form.php
get clients, install and run them inputing your user info.
http://climateapps2.oucs.ox.ac.uk/cp...c/download.php

Then go here and join (top of page little link)
http://climateapps2.oucs.ox.ac.uk/cp....php?teamid=62

[Welnic]

The windows version has a built in service installer. I'm not sure that it works on 2000, though. I installed it on a bunch of XP boxen, but I think there was some issue on the one 2000 box I tried. Or it could have be NT, instead.

[Shaktai]

Originally posted by Scotttheking:
Soo, how do I run this?
I've got OSX 10.2.8, Linux, win98 (NOT MY FAULT), and win2k. Win2k of course needs to be CLI so I can make it a service.

Welcome aboard Scott, how's school going?

We are in a position, to make a run on the #2 position, and with a bit more power, #1 is not out of the question. mikkyo's right though, so far no one has been able to compile a really stable 10.2.x client. Not impossible, but it will probably take some work. For Win 98, make sure you are using BOINC 4.09 or later. Linux and Win 2k should be fairly straight forward, but I know nothing about running it as a service. Work units take 3+ weeks for my fastest machine, a P4-3.4ghz, but it trickles completed work for points 4-5 times a day. Unlike other BOINC projects, it does not require "validation" by another user, so you don't wait for your credit. It is also a worthwhile scientific cause.

[Scotttheking]

Originally posted by Shaktai:
Welcome aboard Scott, how's school going?

1:45am, I just got home, class at 9am. It's busy! I should sleep.

[mikkyo]

I think #1 is possible, those Czechs are putting up a fight for 9th though.
They have added several members in the last week.

Our 32nd member should help quite a bit

[mikkyo]

See what happens when I spend time working on new clients???
Shaktai and Welnic leave me in the dust, that is what.
Next thing cobra will get all his machines online and blow passed me too.
Just keep cranking so we get to #1 and I'll be happy

[mikkyo]

Those Czechs just won't give up easy!
They have added many members and tons of production and machines over the last 2 weeks and are still barely staying ahead of us.
We just need a little more power to eek passed them.

The good news is we are nearly to the point where both of us will overrun UniBE.CH - SwissTeam.NET and OcUK - Overclockers UK.
Now would be a good time for a little push.

If you are also Boincing Seti, you can temporarily focus on CP by changing your Seti Resource share to 20000 and CP to 1.
To work only on CP for awhile, stop boinc, and move your account_setiathome.berkeley.edu.xml file to tempaccount_setiathome.berkeley.edu.xml, and restarting boinc, this will prevent any Seti work form being done at all, until you remove the temp from in front of the account name.

[mikkyo]

We are now in 9th place, 8th should come within the day.
The Czech National Team is still ahead of us, but wait!
Here comes cobra ramping up around the outside of the final turn..
Thinks are looking good, we are presently 3rd in RAC and with all the recent drive the completed units and trickles are just starting to come in.
If anyone has machines they can crunch over the weekend, now would be a great time to get them going.

[Shaktai]

Originally posted by mikkyo:
We are now in 9th place, 8th should come within the day.
The Czech National Team is still ahead of us, but wait!
Here comes cobra ramping up around the outside of the final turn..
Thinks are looking good, we are presently 3rd in RAC and with all the recent drive the completed units and trickles are just starting to come in.
If anyone has machines they can crunch over the weekend, now would be a great time to get them going.

Ya Hoo!!! With Cobra coming on like gangbusters, I am sure to get knocked down another notch, but it is all for the good of the team. #2 spot is very doable. In fact, barring another surge by the Czech National Team, we might move up two spots in the next 48 hours or less. According to BOINC Stats we were actually #2 in total production for the day.