Just an update, I am seeing successful trickle messages being sent to the server as well as intermediate uploads happening. From our testing and what I'm seeing now, uploads are about 100KB each for a total of about 1MB for the entire work unit.

Thanks,
-Uplinger

Sek, The goal here is that results will be truly zero redundancy. And then from that stand point, we will be doing a few things on the backend to both speed up the end results for the researchers as well as make the results run more efficiently.

So, the simulation you are doing is 100k steps. The total steps to complete a given job from the researchers is 2million. So the results from one result will go into the subsequent result. The intermediate uploads are able to act as a way to restart a work unit from those points. Thus we will be validating the work unit as it goes along.

Also, instead of waiting for an entire batch to be completed before we send the results back or start the next stage of a job, the next work unit will be generated automatically by the results that are there.

As you can see this has a level of complexity that we have not attempted in the past, and we will be attempting to have these tests done within beta soon.

In the scenario you provided, the next work unit will be sent out and start from say step 90,000 and attempt to do the next 100k steps. Still having the same 10% checkpoint.

I know, this probably raises more questions than it answers, but more complexity requires more testing :)

Thanks,
-Uplinger

I think I am seeing odd numbers for cpu/clock time

Snagged 5 of them, all on one machine. This will be interesting as that is an 8 core server which is only running off of a 16gb flash drive.
Addendum: 20% after 8 hours 1.86ghz Xeon 5320 Linux Mint 17. Definitely big jobs. Running smoothly so far.

Rats !!! All 5 errored out. The relevant part was:

process exited with code 38 (0x26, -218)...
....
forrtl: No space left on device
forrtl: severe (38): error during write, unit 2, file /var/lib/boinc-client/slots/4/md.out
Image PC Routine Line Source

Each of the WU's errored, but I think this was the cause which made the others error.

I'll post all of the stderr_txt files if needed.

Further update July 31, 2015: After about 9 months of crunching MCM1 on the flash drive with no problems, I think what ever BEDAM is doing may have finally fried the flash drive. It may have been coincidence that it happened at the same time as the beta test, but at any rate, the flash drive has been toasted. I have tried everything I can think of to resurrect it and no luck. It remains impossible to write to it any longer. Another US $6.00 down the drain (sarcasm). I will install another drive and maybe get in on the next beta.
Cheers

Longest running on desktop 11:25 hours at 51% and 11:41 on laptop at 38.2%... with that, BOINCTasks logged 38 checkpoints, just as prescribed. Unfortunately, the standard trickle message does not tell against which task the progress is made, so tried the debug function.

7/30/2015 5:39:01 AM [trickle] read trickle file projects/www.worldcommunitygrid.org/trickle_up_BETA_avx101118-005_r5_1fn_1_1438227471.xml

Just what I wanted to know, which task is doing a progress report.

But actually, not needed as just before or after there are the uploads either starting or finishing.

7/30/2015 7:26:18 AM Started upload of BETA_avx101118-005_r5_1eg_1_3
7/30/2015 7:26:18 AM Started upload of BETA_avx101118-005_r5_1eg_1_13
7/30/2015 7:26:18 AM [sched_op] Starting scheduler request
7/30/2015 7:26:18 AM [trickle] read trickle file projects/www.worldcommunitygrid.org/trickle_up_BETA_avx101118-005_r5_1eg_1_1438233977.xml
7/30/2015 7:26:18 AM Sending scheduler request: To send trickle-up message.