Dear all,

We are excited to announce the return of RakeSearch!

Long-term collaboration with our colleagues from other Russian cities and the vast space of unsolved combinatorial problems have given us new challenges to be solved. The first challenge is to beat the current record on the number of transversals for DLSs of order 12. We already have tasks for the next 2 months, and there will be more.

Technically, we have moved the project to the new hardware server and updated the BOINC server software. We hope that this new setting will bring you a good experience!

With best wishes,
Natalia and hoarfrost

Some workunits from first bunch of ~170 items may fail with an error. We changed the attributes of workunits and expect that new tasks for same workunits will compute successfully.

Hi all,

I've tried some work units so far. Is 3 seconds per workunit the average time ?

I've had some compute error as well. The workunit going to 100% in a dozen of seconds then failed after 5 minutes.

Just a report of what I've observed so far :)

Best regards,

micropro

Hello, micropro!

Tasks run time varies from several seconds to several hours. Another small bunch of tasks added!

So this is a Windows only project now?

Hello stoneageman!

Only for Windows version of this application available at this moment.

Even though the server status page shows ODLS 10 Wus available (and apparently being crunched by others), all I'm getting are beta WUs, which complete in 5-10 seconds. I have the application enabled in my preferences. Is the server page not keeping up with things :).

(3 minutes later) NVM. getting WUs now. that was weird.

Hope you're going to keep awarding the cool badges for the new project. :)

lets start crunching :)

My first task shows completed successfully but the Task detail throws and error at the beginning.... Error: startup_info.hStdInput is invalid.

See below

<core_client_version>7.16.20</core_client_version>
<![CDATA[
<stderr_txt>
20:30:36 (29116): wrapper (7.9.26016): starting
20:30:36 (29116): wrapper: running rake_odls12_windows_intelx86_64.exe ()
Error: startup_info.hStdInput is invalid
11/5/2021 8:30:45 PM Resources freeing: XmlReader
11/5/2021 8:30:45 PM Spstarter finished successfully
20:30:46 (29116): rake_odls12_windows_intelx86_64.exe exited; CPU time 0.000000
20:30:46 (29116): called boinc_finish(0)

Thanks
Bill F

---

I have 40 WUs that errored out and all were longer running tasks (~12,000 sec). Just saying.

The run time for these results are all over the place.

Most runs only seconds, others run long and some error out for running to long.

I think some settings need to be looked at.

+1

Will there be checkpoints when crunching new tasks?

---

Thank you for report! We have raised the border 10 times more.

too many little writes not good for ssd storage

Hi folks!

Thank you for report! We have raised the border 10 times more.

If you have any tasks fallen with and error like "exceeded elapsed time limit" please report about it. Tasks received after my quoted message should not drop out with such an error.

---

Will there be checkpoints when crunching new tasks?

Hello! :) Unfortunately, the current application does not make a checkpoints.

Thank you for participation!

What for checkpointing, it is problematic to implement checkpoints in this experiment, because the sub-trees on which we perform exhaustive search are very different in sizes. If we do a checkpoint at higher tiers of the tree, it is very likely that 9 of 10 sub-trees will be processed very fast, and the 10th will take several hours, so no sense in a checkpoint. If we go into the recursion 1-2 tiers deeper, the check for need of a checkpoint will be performed too often and increase the whole runtime of a task several-fold.

What for checkpointing, it is problematic to implement checkpoints in this experiment, because the sub-trees on which we perform exhaustive search are very different in sizes. If we do a checkpoint at higher tiers of the tree, it is very likely that 9 of 10 sub-trees will be processed very fast, and the 10th will take several hours, so no sense in a checkpoint. If we go into the recursion 1-2 tiers deeper, the check for need of a checkpoint will be performed too often and increase the whole runtime of a task several-fold.

How about using a fixed time interval for a checkpoint trigger with the checkpoint data pointing to the start of whatever calculaton that is happing at that moment?