LHC@home

Yea they need to increase that deadline if thats the case, not everyone can run these 24/7.
Also fix the progress bar if possible.

Both of my computers have downloaded over 500 WUs. I canceled some of them and reduced the cache for WUs to 0.1 days.

---

I have over 100 finished at dev that took 4 or 5 days each a week ago and just sitting there since no wingman have run them

I have 2 tasks that have 12 hours left until the deadline. They ran non-stop for over 5 days but I'm not sure they'll actually finish on time...

---

Both of my computers have downloaded over 500 WUs. I canceled some of them and reduced the cache for WUs to 0.1 days.

how come? What did you set under "max. number of tasks" in your settings ?

unlimited I assume

Both of my computers have downloaded over 500 WUs. I canceled some of them and reduced the cache for WUs to 0.1 days.

how come? What did you set under "max. number of tasks" in your settings ?

It was set to the default setting of 500.

---

I can run these probably 4 - 5 hours a day tops on this machine so its confirmed they will take 4 - 5 days straight on a 6 day deadline ill cancel them all off now and save some power.

unlimited I assume

No, this is the default setting for new machines at LHC@home. :(

---

Hi all,

Thanks a lot for your feedback, and apologies for the bad time estimates.
What we have tried to address now are the extremely short jobs, which were a known pain from SixTrack.
We don't (yet) provide time estimates ourselves when submitting the job to the BOINC server, so I guess there is an automated script that estimates it (and does so poorly).
We will iterate with our IT team to do this better by providing our own time estimates, which we should be able to get (for normalised CPU time) based on the job itself.

A bit of context on our typical job.
An Xtrack job tracks particles through an accelerator lattice (typically LHC).
The CPU time is defined by the product of three variables:

- The number of particles (typically a few hundred to thousands)
- The number of elements in the lattice and their complexity (typically 10'000-100'000)
- The number of turns to track (ranging from 100 to 1'000'000)

This is, of course, only an average estimate; particles can die at any time during the job, when they grow to large amplitudes and hit the beam pipe, where they are no longer tracked.
This was the issue with the SixTrack jobs: particles were sampled on a radial transverse grid and split as such into different jobs.
Some jobs would only contain particles at very high initial amplitudes and die in the first few turns, leading to jobs of literally seconds, and others would be in the core of the beam, surviving the full tracking.
Now we distribute them more evenly, so jobs should have similar running times.

On a recent CPU, one should get around 1000-2000 particleturns / second for the lattice used in the jobs that are being submitted now.
This means that we expect these jobs (100 particles and 1 million turns) to take 14-28 hours.

However, we had an earlier batch of jobs where we made a calculation mistake.
These jobs take many days to finish, and most of them will, unfortunately, hit the max runtime.
We have cancelled the submission (because otherwise they will keep on being resent), but we cannot cancel them on your side.
Feel free to cancel them to free CPU time.
For ease of recognition, we have marked new jobs (with running time of 14-28 hours) with "_ScanShort" in the name for now, so these ones you'll want to keep.
Our sincerest apologies for this hiccup.
We are working to get the system as robust as possible with as few inconveniences for you as possible, but it is a work in progress.

As always, thanks for crunching our jobs.
Your contribution is highly valued.

Cheers,

Frederik

Hello Frederik,

Thanks for update.

---

Server has cancelled all my XTrack tasks.

---

Server has cancelled all my XTrack tasks.

Everything seems to have been cancelled both here and in the development version. Let's hope they'll fix the runtime issue and then we can take another swing at it.

---

Hello Frederik,

Thanks for update.

---

Supporting BOINC, a great concept !

https://lhcathomedev.cern.ch/lhcathome-dev/results.php?hostid=5072&offset=0&show_names=0&state=5&appid=16
https://lhcathomedev.cern.ch/lhcathome-dev/results.php?hostid=4786&offset=0&show_names=0&state=5&appid=16
https://lhcathomedev.cern.ch/lhcathome-dev/results.php?hostid=5326&offset=20&show_names=0&state=5&appid=16

Thanks for the update

I have three jobs 'completed' and 'validated'.
Runtime was roughly 2:30h.
Three out of xx.xxx. :(

Server has cancelled all my XTrack tasks.

5 tasks are cancelled for me. The others are still running, but how long?
I'll wait till it's more reliable.

Server has cancelled all my XTrack tasks.

5 tasks are cancelled for me. The others are still running, but how long?
I'll wait till it's more reliable.

Only the new batch of tasks with "_ScanShort" in the name should be kept. Everything else should be aborted.

---

Server has cancelled all my XTrack tasks.

5 tasks are cancelled for me. The others are still running, but how long?
I'll wait till it's more reliable.

Only the new batch of tasks with "_ScanShort" in the name should be kept. Everything else should be aborted.

Scratch that, they got cancelled again.

---

all mine just server aborted. didn't go through all of them but many were named like so-


camontan__bbb_end_of_levelling_scan_1005__ScanShort_10_3__VvxcKlckpnN5

camontan__bbb_end_of_levelling_scan_1005__ScanShort_34_4__AzjhN7YzbZfl

---