1/6-8/6 Citizen Science Grid Server Stress Test for DC-Vault

Alez · Post by **Alez** » Mon May 25, 2015 1:53 pm

http://boincstats.com/en/stats/challenge/team/chat/685

Name Citizen Science Grid Server Stress Test for DC-Vault possible inclusion
Status Upcoming
Project Citizen Science Grid
Issued by [H]ard|OCP
Start time 2015-06-01 00:00 UTC
End time 2015-06-08 00:00 UTC
Late entrants allowed? Yes
Number of teams participating Â» 10
Number of users participating 0

Alez · Post by **Alez** » Wed May 27, 2015 12:48 pm

Going to get set up for this one. Might even make an attempt at a bunker if I remember

Anyone else fancy it ?

Silver · Post by **Silver** » Wed May 27, 2015 1:09 pm

I've never run it before, what's the deal with this one Alez.
Cpu and or gpu? Any hints it tips

Post by **scole of TSBT** » Wed May 27, 2015 2:05 pm

I'm in.

As far as I know it's CPU only, but they did recently announce a droid app for SubsetSum@Home. Just attached a droid to csg and have a few WU. Will see how it does.

Current deadline for WUs is 05/31 so just a little too soon to bunker. How long their WU stay pending? Is this a project where it helps to build pending before a challenge?

IP address range to block for bunkering 134.129.182.0 --> 134.129.182.255

Alez · Post by **Alez** » Wed May 27, 2015 3:39 pm

Yip no GPU app. Going for pending just now, will try for a bunker on the last day I think. Let us know how the droids do and I will attach mine on as well once I have reconnected them again .....

Post by **scole of TSBT** » Wed May 27, 2015 11:37 pm

DNA@Home pays a little better than SubsetSum@Home. The droids do really well. I only posted all the detail to show how well the droids do.

Avg Run Sec and Avg Cr are of limited value because of different WU sizes.

App________________System___________Avg Run Sec___Avg Cr__Avg Cr/Sec
DNA@Home_________xeon x5570 2.93ghz__1631.6______12.2_____0.0075
DNA@Home_________i3-3240 3.4ghz_______1104.6______8.9_____0.0081
DNA@Home_________phenom II 3.4ghz______981.6______8.0_____0.0082
DNA@Home_________xeon x5650 2.67ghz__2937.8______23.7_____0.0086
DNA@Home_________xeon E3-1246 3.5ghz__2094.1_____19.8_____0.0096
DNA@Home_________i7-3770k 3.5ghz_______740.9______8.3_____0.0112
DNA@Home_________i7-3770 3.4ghz_______3596.6_____41.1_____0.0115

SubsetSum@Home___odroid U2 1.7ghz_____17409.0_____57.0_____0.0033
SubsetSum@Home___rk3188 1.6ghz_______14058.1_____59.2_____0.0042
SubsetSum@Home___s802 2.0ghz_________10367.6_____56.6_____0.0054
SubsetSum@Home___xeon x5570 2.93ghz___8105.4_____57.6_____0.0071
SubsetSum@Home___i3-3240 3.4ghz_______8551.9______63.8_____0.0075
SubsetSum@Home___rk3288 1.8ghz_______7276.9______55.4_____0.0076
SubsetSum@Home___i7-3770 3.4ghz_______6599.0______57.3_____0.0087
SubsetSum@Home___xeon E3-1246 3.5ghz__6598.9______58.7_____0.0089

Had limited results before. Updated the rk3188 numbers. Still good.

Alez · Post by **Alez** » Thu May 28, 2015 1:49 am

Excellent, going to get all my droids on this to. Lets see what we can do .

Post by **scole of TSBT** » Thu May 28, 2015 5:28 pm

Well the project admin is taking all fun out of this by setting the WU in progress limit to # of cores + 1. I know how to change <ncpus> to fool a project, but how do you do change <ncpus> to 1000 without BOINC launching 1000 tasks? Changing anything to limit the number of active tasks counters the increased <ncpus> and no more WUs will download. At least that's what I'm seeing.

Janos (retired) · Post by **Janos (retired)** » Thu May 28, 2015 6:04 pm

Not tried this myself but I would posit you could retain the status quo so that as you increase <ncpus> then "limit" your CPU usage % by the reciprocal.

Post by **scole of TSBT** » Thu May 28, 2015 6:15 pm

Yeah, I tried that too but it still launches the same number of tasks as <ncpus>, which doesn't overwhelm the cpus, but it does use all your memory and then starts paging it out, which hoses the system. The DNA WUs can use over 500MB.

Alez · Post by **Alez** » Thu May 28, 2015 6:58 pm

change ncpus to a ridiculous number, update the project and download tons. Suspend boinc. Change ncpus back to the real number, re-read config file and unsuspend boinc. ( might have to shut boinc down and then restart boinc. You should now be crunching the correct number of cores and have a massive bunker.

I'm going to be away until late sunday so looks like I will just have to go the pending route rather than bunker

Post by **scole of TSBT** » Thu May 28, 2015 7:54 pm

Won't it start launching tasks? How many 500MB WUs will launch before your system becomes fatally unresponsive?

Post by **scole of TSBT** » Thu May 28, 2015 11:27 pm

Can't bunker and pending levels off quickly so I was going to back off until it starts, but we've moved CSG to 12th in the Formula BOINC standings so I'll stay with it. :)

Alez · Post by **Alez** » Fri May 29, 2015 12:24 am

scole250 wrote:Won't it start launching tasks? How many 500MB WUs will launch before your system becomes fatally unresponsive?

They can't launch because they will end up in a waiting for memory status. Once you set ncpus back to normal and re read the config file then it will only launch the numbers of cpu's as per that. I also believe that once a unit is in a downloading status it will complete it if you suspend then allow boinc to run again. Not 100% as I've never tried it, but it should work ( in theory ).
I'm cracking on with it to, both for formula boinc and get as many pendings as I can because I will not be able to attempt to bunker.
Give it a try if you wish and let us know if it works or not.

Post by **scole of TSBT** » Fri May 29, 2015 1:12 am

How many 500MB WUs will launch before your system becomes fatally unresponsive?

On a 8 core i7-3770K w/16GB RAM and CPU usage set to .5%, 39 WUs before it became fatally unresponsive. Try at your own risk. Remember, once the physical memory is used it will start paging to disk, so you'll need enough disk space (and patience) to page out all those active tasks where many (most?) tasks can be over 500MB.

Alez · Post by **Alez** » Fri May 29, 2015 3:02 am

Once it starts downloading the units you then suspend boinc and reset the ncpus back to 8, do advanced , read config files. If you do that does it not complete all the downloads but only starts 8 tasks ?
That's how it used to work.

Post by **scole of TSBT** » Fri May 29, 2015 11:19 am

In order to change the <ncpus> value, you must re-load the cc_config.xml which requires restarting the BOINC client and that kills the downloads in progress.

Post by **scole of TSBT** » Sat May 30, 2015 11:53 am

Well, I used the app_config.xml to limit the number of active CSG tasks and it's working. I thought I tried that before and it wouldn't download

Oh well. Project still only allows 256 per system max, which may not be enough for the 8 thread i7 3770 systems but I'll bunker them now anyway.

Post by **scole of TSBT** » Sat May 30, 2015 4:39 pm

Boinc Tasks allows you to edit the cc_config,xml on any connected system, such as androids. You can change the <ncpus>, re-read the cc_config.xml and it works to download a bunker. I thought the BOINC Client had to be restarted to reload the cc_config.xml, eh. It will launch as many tasks as CPUs you've set, so do as Janos suggested and set the "Use at most % cpu time" to something low, like .5. I had 80 active tasks and it didn't crash. Just set the <ncpus> and % usage back when you have all the WUs you need. Anyway, that's a method to bunker this project on androids.

Alez · Post by **Alez** » Sat May 30, 2015 9:25 pm

scole250 wrote:I thought the BOINC Client had to be restarted to reload the cc_config.xml, eh.

..... as I pointed out ' do advanced , read config files ', most things can be changed in the boinc manager now on the fly. Some still need a restart such as changing the number of active miners. Won't be home till late tomorrow so no point in me bunkering. Will be a standing start effort which should improve once i get a few systems back up and running.

Post by **Megacruncher** » Sun May 31, 2015 1:58 am

It's a bit last minute, and I'll certainly not be doing any bunkering, but I'm now signed up for this project & I'll do what I can for the team in this challenge.

Alez · Post by **Alez** » Mon Jun 01, 2015 12:44 am

Windows has broken my xw6600 server I use to control all the others or the miner blowing up has fried the usb chip in that too as it no longer recognises any usb device ( mouse , keyboard etc. ) though it claims driver errors so not sure if the hardware is bad or windoze is fried. Will have to get a pci - usb card to determine.

Alez · Post by **Alez** » Mon Jun 01, 2015 3:07 pm

Now the dust has settled from the initial start TSBT is sitting in 3rd, a position we can hopefully hold. scole has assumed his usual position at the head of the pack in 1st but it might be tough to hold off grandpa .
I'm in 10th, paul 30th, skywalker 31st and mark in 57th. Well done us and crack on

Silver · Post by **Silver** » Mon Jun 01, 2015 6:28 pm

A great start indeed chaps!
Keep up the good work

Post by **Sphynx** » Mon Jun 01, 2015 7:58 pm

Put this in the app_config.xml under the project:

<app_config>
<app>
<name>Citizen Science Grid</name>
<max_concurrent>8</max_concurrent>
</app>
</app_config>

Change the <ncpus> to whatever you want in the cc_config.xml. The project will see the spoof processors but you will run no more that 8 wus at a time. If it's a quad, change the 8 to 4, hex to 6. Regardless of what you're running, DNA or SSS, it will only run the number of <max_concurrent> you specify. Only works on new versions of BM and prevents system lock.

@scole - nice bunker!

Post by **scole of TSBT** » Mon Jun 01, 2015 8:13 pm

I had tried that late last week and thought the project was still holding back WUs unless you allowed all the CPUs to launch a WU. It of course crashed my system. I think I was just impatient because it seems to take a long time for WUs to download and upload to the server. I did try it again in time to build a little bunker. You know what they say down here along pit road...if you ain't cheating, you ain't trying hard enough

Post by **Sphynx** » Mon Jun 01, 2015 8:16 pm

Sorry about the repeat post. DNA uploads have my network all messed up.

Post by **scole of TSBT** » Tue Jun 02, 2015 1:28 am

Sphynx wrote:Sorry about the repeat post. DNA uploads have my network all messed up.

Slow day on the forum. Gives us more posts to read.

Alez · Post by **Alez** » Tue Jun 02, 2015 3:10 am

Sphynx wrote:Sorry about the repeat post. DNA uploads have my network all messed up.

What multiple posts are you referring to ?

Mega up on the challenge now in 47th and rising, anyone else likely to pop up ?

Much as we are always happy to see you guys at seti.usa around in a challenge, does this mean 4th is imminent for us ? Guess we will have to try and make that as difficult as we can for as long as we can

Post by **Sphynx** » Tue Jun 02, 2015 12:50 pm

Alez wrote:Much as we are always happy to see you guys at seti.usa around in a challenge, does this mean 4th is imminent for us ? Guess we will have to try and make that as difficult as we can for as long as we can

I don't think you have to concern yourself with SUSA. We look to be a 6 place finisher...maybe 7th if we're not careful.

Alez · Post by **Alez** » Tue Jun 02, 2015 5:19 pm

Seems the credit allowance for DNA has been drastically cut to credit screw levels. Another project marred by lousy returns for your efforts.

Also seems that the credit realignment has broken the credit count on the competition.

Post by **scole of TSBT** » Tue Jun 02, 2015 7:57 pm

Well all I know is if Travis, the CSG admin, fixes everything and BOINCStats gets the challenge stats corrected, I'm going to demand my bribe back!

Challenge total____723,320
Actual total_______284,142

Silver · Post by **Silver** » Tue Jun 02, 2015 8:23 pm

I'm not sure what just happened but I feel Sepp Blatter must've had something to do with it

Alez · Post by **Alez** » Tue Jun 02, 2015 8:34 pm

Silver wrote: I'm not sure what just happened but I feel Sepp Blatter must've had something to do with it

There's no corruption in the FIFA family ...... the mans a

Post by **Dirk Broer** » Wed Jun 03, 2015 12:52 am

The only stress that is being tested is the stress of the volunteers!

Post by **scole of TSBT** » Wed Jun 03, 2015 12:55 am

Well I'd be happy if they would just let the current credits stand and fix it going forward

And I have 11 droid cores sitting, waiting for work. Hate to turn backup project loose just yet.

Alez · Post by **Alez** » Wed Jun 03, 2015 2:29 am

I've lost shed loads of credit, somewhere close to 250k ( 264,174 to be precise ) and those of us that were top 10 in the challenge now seem to be in complete limbo. Bit of a fiasco really.
and also the lack of subset work is starting to hurt as DNA units can take up to 3/4 GB of memory each, I've now got systems crashing as they run themselves into the ground. Uploads are real slow to and starting to back up which isn't helping either.
Not going to stick with this much longer before I go back to other projects that don't jerk us around halfway through a competition. Not a clue where either myself or the team stands in this now

My 10 droids ( all quads ) have already picked up other projects and I have no intention of forcing them back.

Dirk, seems the word volunteers is lost on a lot of projects recently ......

Alez · Post by **Alez** » Wed Jun 03, 2015 12:37 pm

Stats still all down for both teams and individuals although a few do seem to be updating. Mega and Paul seem to be getting credited, everyone else dead in the water. Lost all the points bar 2 on formula boinc to.
Travis proposing a fixed credit system http://csgrid.org/csg/forum_thread.php?id=2073
Feel free to chuck in your two pence worth

Post by **scole of TSBT** » Thu Jun 04, 2015 1:55 am

Looks like Grandpa will overtake me for 1st. Who knew the old guy could move so fast

On the bright side, looks like I'll earn another WUProp magenta app. One step closer to that next star

Alez · Post by **Alez** » Thu Jun 04, 2015 2:06 am

Seems grandpa was the only one in the top 10 who reset. Not sure whether to believe any of the stats on the competition now or not. Very few people seem to be moving.

Just finished replacing yet another PSU on my miners. A 500w CIT unit running a single R2 box and the plug to the miner has melted totally. Saved the miner but the PSU is toast. Sure that's the last CIT unit I have. Will check anyway as any others I have are getting binned. Wires from the PSU are far thinner than any other PSU's I have and the R2's don't pull that much amps. Cheap, cheap crap.

Post by **scole of TSBT** » Thu Jun 04, 2015 11:57 pm

What do you think gang? This challenge is screwed up worse than soup sandwich. Do we stay on it until the end or go focus on other team goals? I'll stay with it if that's the popular opinion. Just axin'.

Post by **Megacruncher** » Fri Jun 05, 2015 12:26 am

I'm going to stay in but only because I'm still moving up the individual ranks. This is because I'm a newbie here and wasn't too badly hit by the realignment. Basically I didn't have that many credits to lose. Some of the rest of you are being perversely punished for your prior investment in the project. You might as well cut your loses and bail out

Post by **scole of TSBT** » Fri Jun 05, 2015 12:44 am

I'll stick with it too. I don't care about my individual spot. Grandpa is killing it and deserves top finish. The only downer for me is as a team challenge, this is over. Stats are FUBAR. Don't blame the CSG project admin though. I don't think he expected this type of disruption. He's very attentive to the project compared to others and deserves support.

Alez · Post by **Alez** » Fri Jun 05, 2015 1:24 pm

I'll stick with it and see if we get any more points on FB challenge, though I will also try and get some of the points we have just lost back on dDM, Yafu and POGS.

Silver · Post by **Silver** » Fri Jun 05, 2015 4:33 pm

I'll stick with it to the end too.
Nothing to do with the fact I'm really lazy and can't be bothered changing

Alez · Post by **Alez** » Wed Jun 10, 2015 11:44 am

Nothing really to report for the end of this as the credit system was completely borked. Never the less we did very well, gained points on formula boinc, and gained 20 places on bam though the credits still seem to be bouncing around. When the new credit system is implemented then I guess it will all change again.