The SETI.Germany vs. TeAm AnandTech race in Correlizer. The goal has been reached!!!

[wayliff]

server must be getting pounded...

here is a response from the server admins at the Correlizer message board:

"I'm sorry for the long delay, there where some internal things that needed to be tackled in the Netherlands. I've added multiple instances of the validator and assimilator to cope with the workload, readjusted some kernel and system settings so at least the tasks can be fetched. We are also planning to setup a multi host environment in the next couple of weeks.

Anis Abuseiris
Erasmus Grid Office"

http://svahesrv2.bioquant.uni-heidelberg.de/correlizer/forum_thread.php?id=51#354

 

[wayliff]

I think there's something I've noticed...

If you have many items to report, it can take a long time for the reporting to actually occur.

Example: Three of my machines had 30+ to report and yet were unable to report but another had only 2 and I noticed it has been uploading without an issue. (I still have one machine that has dozens of completed WUs and are not reported yet)

Perhaps we need to increase the amount of times the boinc client talks to the server to do smaller but more often transactions.

Opinions?

 

[Sunny129]

the Correlizer server is just all f_cked up right now. according to BOINC, of all the Correlizer tasks that are "in progress" on my host, their deadlines are either 11/18/11, 11/24/11, or 11/25/11. and yet my account on the Correlizer server is showing that i have tons of v1.07 tasks "in progress" that are past their deadline of 11/11/11, ~9:30 UTC, but haven't crunched yet. what's even more strange is that the issue date on all of these tasks is 11/11/11, ~11:00 UTC!

obviously something's funky, b/c the server does not send out tasks and expect them to be crunched and returned in just ~90 minutes lol! everything looks good on my end (the BOINC client), so i'll just wait for them to sort it out on the server end...i'm sure the task stats will sort themselves out soon enough...

 

[geecee]

I think there's something I've noticed...

If you have many items to report, it can take a long time for the reporting to actually occur.

Example: Three of my machines had 30+ to report and yet were unable to report but another had only 2 and I noticed it has been uploading without an issue. (I still have one machine that has dozens of completed WUs and are not reported yet)

Perhaps we need to increase the amount of times the boinc client talks to the server to do smaller but more often transactions.

Opinions?

Interesting, I've noticed the same thing. The larger the reporting backlog, the more issues it has completing. My X6 box had 45+ WUs to report this morning and kept getting gateway errors. OTOH, my relatively slow dual Xeon work box, which only goes through a few WUs every 3-4 hours, had no problem reporting its work as far as I can tell, at least nothing in the BOINC msg log. I finally got the X6 to report, but it took many tries. Interestingly, after it finally reported the large number, the next couple of units it reported didn't seem to have any issues getting through. Strange.

 

[Uppsala9496]

11/11/2011 10:07:11 AM | correlizer | No tasks are available for Correlizer Applications

That's from my log file. My queue is getting short again on my X6.

 

[wayliff]

11/11/2011 10:07:11 AM | correlizer | No tasks are available for Correlizer Applications

That's from my log file. My queue is getting short again on my X6.

The server is truly out of whack...I suppose...I just reported 16 on one machine and got back 9 new WU.

it must be frustrating for sure...

 

[Plimogz]

Odd, the server was showing something like 17K tasks ready to sent out, a little while ago.

 

[Uppsala9496]

We broke them

Waiting to report 143 tasks on one rig; 117 on another.
Curious to see how this impacts that RAC over the next few days.

 

[Plimogz]

have you tried updating from the project tab? It seems to help me clear up those "ready to report" wu's. I got the impression that it also helps get new work. I think that both "ready to start" and "ready to report" tasks count against the work buffer

 

[Uppsala9496]

I did update them. My daughter is watching Sesame Street right now, and the HTPC is hooked up to the tv and is remote desktop controlling the other rig. I quickly checked (and was told not to change the channel) and hit update a few times for both machines. We'll see what happens. Both have pretty good queue's of work unlike my X6.

 

[petrusbroder]

I have more than 450 WUs "Ready to report". I am getting these reports:

2011-11-11 18:07:09 | | Project communication failed: attempting access to reference site
2011-11-11 18:07:10 | | Internet access OK - project servers may be temporarily down.

It will clear up sooner or later ...

 

[blckgrffn]

I have added a couple CPUs... check my computers tab Calling in the reinforcements.

Hopefully I have some sort of score by Monday...

 

[mk]

I have more than 450 WUs "Ready to report". I am getting these reports:

It will clear up sooner or later ...

Same thing here, 467 WU's to report and counting. Hopefully they'll get through in the not so distant future. Me want credit!

 

[Sunny129]

I have added a couple CPUs... check my computers tab Calling in the reinforcements.

blckgrffn's machines

very nice...thank you very much for your contribution to the project! :thumbsup:

btw, are the three E7450 machines one and the same and haven't been merged yet, or are they 3 individual machines?

 

[blckgrffn]

Big virtualization boxes idled between projects, 4 way Dunnigtons w/256GB of RAM.

So, each are individual sockets

There are a couple more now too, but I am getting errors

 

[bryanW1995]

I have more than 450 WUs "Ready to report". I am getting these reports:

It will clear up sooner or later ...

Ha, we should tell them to try out SETI sometime, right? A couple of days' outage with SETI here or there is nothing.

 

[wayliff]

I have added a couple CPUs... check my computers tab Calling in the reinforcements.

Hopefully I have some sort of score by Monday...

Thanks for all that help!
It should start to crunch nicely!

Did you just add most of these very recently? Do you have all machines crunching WUs? Reason I ask is because on the site most of these machines have very little or no WUs.
Good to have you on board with that small platoon at your command!!

Anyway, SG has new correlizer member ('pan2000') that has some serious PrimeGrid crunching numbers, or at least it looks like a high number to me. I don't know if that is all from GPU but I am guessing this member is heavy artillery.

 

[Sunny129]

Anyway, SG has new correlizer member ('pan2000') that has some serious PrimeGrid crunching numbers, or at least it looks like a high number to me. I don't know if that is all from GPU but I am guessing this member is heavy artillery.

yeah, his computers are hidden, but 170,966 avg. credit in PrimeGrid i believe is on par with a day's worth of output from a single 5870 GPU...or like you suggested, he could have some major CPU power at his disposal...i'm not getting my hopes up yet, just in case those PrimeGrid numbers are from a GPU and not CPUs...

...either way, we welcome him to the TeAm!

 

[wayliff]

...either way, we welcome him to the TeAm!

absolutely!

 

[bryanW1995]

yeah, his computers are hidden, but 170,966 avg. credit in PrimeGrid i believe is on par with a day's worth of output from a single 5870 GPU...or like you suggested, he could have some major CPU power at his disposal...i'm not getting my hopes up yet, just in case those PrimeGrid numbers are from a GPU and not CPUs...

...either way, we welcome him to the TeAm!

Most people with very high numbers in any project are gpu-focused, so don't be too worried about one member here or there from SG keeping them ahead. People like Petrus with eleventy billion cpus are best-suited for a project like correlizer.

Looks like we are making steady progress since the cpu forums thread was posted. A lot of those guys are pumped about it, and it's especially nice for the AMD guys to feel appreciated after all that BD mess...

 

[waffleironhead]

We outproduced SG yesterday!

TeAm=206.362
SG =177,396

Way to go everyone!

 

[blckgrffn]

Thanks for all that help!
It should start to crunch nicely!

Did you just add most of these very recently? Do you have all machines crunching WUs? Reason I ask is because on the site most of these machines have very little or no WUs.
Good to have you on board with that small platoon at your command!!

Anyway, SG has new correlizer member ('pan2000') that has some serious PrimeGrid crunching numbers, or at least it looks like a high number to me. I don't know if that is all from GPU but I am guessing this member is heavy artillery.

Correlizer says the machines are checking in, but they have no tasks So I am guessing that means no WU's? I am probably done spinning up VMs today, but I could probably add another 30 cores next week if they have work to do, in addition to the 30 cores I spun up today.

 

[wayliff]

Correlizer says the machines are checking in, but they have no tasks So I am guessing that means no WU's? I am probably done spinning up VMs today, but I could probably add another 30 cores next week if they have work to do, in addition to the 30 cores I spun up today.

Thanks for that big contribution.

I guess a couple of things to try to get some crunching going.

1. On the boinc client select the project tab, click on correlizer and then update. (I realize you have plenty of clients!)

2. Also you can increase the number of times the boinc client talks to the server...I set mine to ~8 times a day. Launch Boinc client, look for the preferences based on your version, select the network usage tab and put 0.13 in the "connect every # days" field, if that value is already resulting in more than 8 calls to the server, I guess leave it alone...you can also increase the buffer if you'd like.

I had mine set to 4 times (0.25) a day but decided to change it because of the server problems I've seen. In that way my machines report more often and smaller chunks of data.

Hope that helps!

 

[blckgrffn]

Thanks for that big contribution.

I guess a couple of things to try to get some crunching going.

1. On the boinc client select the project tab, click on correlizer and then update. (I realize you have plenty of clients!)

2. Also you can increase the number of times the boinc client talks to the server...I set mine to ~8 times a day. Launch Boinc client, look for the preferences based on your version, select the network usage tab and put 0.13 in the "connect every # days" field, if that value is already resulting in more than 8 calls to the server, I guess leave it alone...you can also increase the buffer if you'd like.

I had mine set to 4 times (0.25) a day but decided to change it because of the server problems I've seen. In that way my machines report more often and smaller chunks of data.

Hope that helps!

I can't find that setting in my BOINC client (the BOINC Manager from the Ubuntu software store) - but all of the big boxes look to be dry while the little ones have plenty to work on for a while. Bummer.

It does look like they are regularly - maybe once an hour at the minimum - checking the site for work...

 

[Uppsala9496]

I can't find that setting in my BOINC client (the BOINC Manager from the Ubuntu software store) - but all of the big boxes look to be dry while the little ones have plenty to work on for a while. Bummer.

Tools -> Computing Preferences -> Network

I'm running Boinc 6.12.33 in Ubuntu 11.10 (straight from the repositories).