Project offline the last 4 days due to router malfunction / Proyecto fuera de linea los ultimos 4 dias debido a averia en router  |
|
On day 5, November, our router broke and stopped working. It has been necessary to replace it. For this reason Falua has been offline. We are online again.
/
El pasado dia 5 nuestro router dejo de funcionar. Ha sido necesario reemplazarlo. Por este motivo Falua ha estado desconectado. Ya estamos operativos otra vez.
|
|
| DNETC@home down (switch off) |
|
Zadecydowaliśmy (OxyOne i Sesef) o zamknięciu projektu DNETC@HOME.
Zapraszam do MooWrapper@home.
Dziękujemy za współprace.
Serwer działa do wyczerpania zapasów WU.
GOOGLE TRANSLATOR
We decided (OxyOne and Sesef) to close the project DNETC @ HOME.
Feel free to MooWrapper @ home.
Thank you for your cooperation.
The server runs while stocks WU.
|
|
| New computations start today / Hoy empiezan nuevos calculos |
|
Today I start testing a new set of WUs. They will be made available in small sets from time to time in the following days
/
Hoy comienzo las pruebas con un nuevo juego de unidades de trabajo. Las ire poniendo en el servidor en pequeñas cantidades durante los siguientes dias.
|
|
| Metastable states. |
| The last run with antialiasing had enough precision to resolve (at least some of) the metastable states. In particular, it allows to build the region of existence of metastable vortices: map_0.5, map_1.0, map_1.5. These three maps correspond to different comparison tolerances (in percents) in the algorithm for metastable state detection (the lower barriers are considered passable). As one can see, the dependence on the tolerance in most regions on the map is rather small to make it sound. This result is new, because the boundary does not completely fit any of the known partial results for the vortex stability line. Right now I'm working on analytical model to supplement this numerical map. There will be no new WUs for some time, probably. This is because in the light of the new developments in the underlying theory the program will have to be significantly modified, to cover much wider array of cases. (moreover, there are even further developments, which I'm readying for publication right now). But, even if there are no new WUs for a while, please consider the project alive. Numerical calculations constitute only a part of it, the smaller part. |
|
| Freehal back again |
Sorry, I was very busy during the last weeks because I'm going to university soon and I needed to look for a flat and I haven't been much at home.
So I didn't had much time for Freehal, and, to be honest, I also needed to take a vacation from Freehal, it's sometimes stressful and there are also other things in life.
Now I can spend more time for Freehal and BOINC again.
--
Best Regards,
Tobias Schulz |
|
| Thank-you all |
|
Dear volunteers,
Your contribution over the last 3 years to the AQUA@home project was critical for us to perform immense computations necessary for the advancement of our science and technology. We have published several scientific papers reporting the results of those computations, including our recent publication in Nature. We made sure to acknowledge your generous contributions in those papers, as your help was vital. At this time, however, we no longer have large computations requiring the computing power of AQUA@home. We therefore have come to the decision to shut down AQUA for the forseeable future. Thank-you again for the selfless devotion of your time, patience, and computing resources to our effort to progress the field of quantum computing.
Sincerely,
Neil Dickson
A list of the scientific papers you've helped make possible:
Investigating the Performance of an Adiabatic Quantum Optimization Processor, in Quantum Information Processing, preprint available at http://arxiv.org/abs/1006.4147
Robust Parameter Selection for Parallel Tempering, in International Journal of Modern Physics C, Volume 21, issue 5, 2010, preprint available at http://arxiv.org/abs/1004.2840
High-Performance Physics Simulations Using Multi-Core CPUs and GPGPUs in a Volunteer Computing Context, in International Journal of High Performance Computing Applications, Volume 25, No. 1, 2011, preprint available at http://arxiv.org/abs/1004.0023
Importance of Explicit Vectorization for CPU and GPU Software Performance, in Journal of Computational Physics, Volume 230, Issue 13, 2011, preprint available at http://arxiv.org/abs/1004.0024
A Performance Comparison of CUDA and OpenCL, preprint available at http://arxiv.org/abs/1005.2581
Quantum Annealing with Manufactured Spins, in Nature, Volume 473, Issue 7346, 2011
Algorithmic Approach to Adiabatic Quantum Optimization, preprint available at http://arxiv.org/abs/1108.3303
|
|
| De eerste 10.000 WUs van de nieuwe versie van Correlizer zitten nu in het systeem |
|
|
| Phase II completed / Completada Fase II |
|
Today we have completed all calculations required for this second set of WU. There will be no more work available until analysis of the results is completed.
/
Hoy hemos terminado los cálculos de nuestro segundo grupo de problemas. No habrá más tareas disponibles hasta que no terminemos el análisis de los resultados obtenidos.
|
|
| Endless WU problem / Tareas que no terminan |
|
A set of wrong WU has been found to produce the worker to never end calculations. We have cancelled all the faulty WU, so if you detect you have one of these (with names "Prueba290000" to "Prueba299999") please cancel it and update to take fresh WU.
/
Un juego de tareas erróneas hace que no se terminen nunca los cálculos. Hemos cancelado todas esas tareas, asà que si detectas alguna (con nombres "Prueba290000" hasta "Prueba299999") por favor cancélala y actualiza el proyecto para obtener tareas nuevas.
|
|
| Asteroid 2011 MD |
The little asteroid 2011 MD, about 10 meters in diameter, will fly by the earth on Jun 27. We have posted here some animations showing the geometry of this encounter. It is interesting to note that it will come closer than the GPS satellites, and also that it comes from the norther hemisphere, passes over the Earth's south pole, and then is deflected so strongly that leaves the Earth back in the northern hemisphere. |
|
| Server fixed, account creation reopen / Servidor arreglado, creación de cuentas abierta |
|
Looks like the problems we suffered the last days have been solved. Our work generator had a bug. We fixed it and raised the number of WU in the cache, so now WUs are flowing as expected. Account creation has been re-enabled. We also upgraded a little bit our HW and have now a double Xeon.
/
Parece que hemos arreglado los problemas acontecidos en los últimos dÃas. Nuestro generador de trabajos tenÃa un error. Lo hemos arreglado y subido el número de tareas en la caché, asà que parece que las tareas fluyen ahora adecuadamente. Por ello hemos vuelto a habilitar la creación de cuentas. De paso hemos actualizado un poco el HW y tenemos ahora un doble Xeon.
|
|
| Server max capability reached / Alcanzada capacidad máxima del servidor |
|
Due to the number of volunteers and hosts that have joined us in the last days, looks like our (modest) server has reached its maximum capability and is not able to serve work units at the required pace. We are forced to disable the account creation by now.
/
Debido al número de voluntarios y ordenadores que se nos han unido, parece que nuestro (modesto) servidor ha alcanzado su máxima capacidad y no es capaz de servir tareas al ritmo debido. Nos vemos forzados a desabilitar la creación de cuentas por ahora.
|
|
| Problem solved, server runing again / Problema resuelto, servidor funcionando de nuevo |
|
This morning, while transferring work units from an external hard disk to the server, for an unknown reason something went wrong. As a result, several thousands WU were copied with length 0. This has the result of producing error "MD5 check failed" and "Checksum or signature error" and, oddly, (not sure if for the same reason or not) the scheduler was not running and the server was unable to keep serving more work. A reboot has been needed (start and stop commands were not able to complete) but now all looks OK. Just wait until the system discards all wrong WU when they reach its fault limit counter. We will copy (with more luck, I hope so) again the WU to the server and reprocess them. Please check if this works.
/
Esta mañana, mientras estábamos copiando unidades de trabajo desde un disco externo, algo ha ido mal (no sé los motivos) y como resultado se han copiado varios miles de tareas con longitud 0. Esto ha producido los errores "MD5 check failed" y "Checksum or signature error" y, lo más raro (y no estoy seguro de que sea por el mismo motivo o no) el planificador ha dejado de funcionar, por lo que el servidor ha dejado de servir trabajos. He tenido que reiniciar todo el sistema (ya que tampoco parecÃa capaz de completar los comandos de start y stop) y ahora parece que todo va bien. Solo hay que esperar a que el sistema descarte las tareas erróneas cuando alcancen su lÃmite de fallos. Copiaremos de nuevo (con más suerte, espero) las tareas que faltan a fin de reprocesarlas. Os ruego que comprobéis si funciona.
|
|
| RAM considerations and Windows bug |
|
As you know, until now, this project has been really RAM demanding. This has led to different kinds of problem/bug:
- If the WU try to consume more RAM than allowed (in BOINC preferences and/or according to the limitation embedded in the WU), an error is usually thrown.
- If the WU try to consume more than 2Go on 32bits platform, an error is thrown. I know there are some ways to avoid this, but you should understand that we do not want to consider solutions (WU) that are consuming more than 2Go. The RAM consumption depends on the parameters embedded in the WU and it's almost impossible to predict the RAM consumption from the parameters. All we do is put some limitation on the value of each parameter individually, but if all the parameters are at their max value, the RAM consumption will of course be huge.
Anyway, I still need to evaluate 10000 WUs (current batch) then I'll stop the experiments on that kind of WU (i.e. no more huge RAM consumption in the future).
Concerning the bug on windows "multiple instances running simultaneously" (see: http://boinc.run.montefiore.ulg.ac.be/evo/forum_thread.php?id=37), I've got some news.
I asked on the boinc projects mailing list, and it turns out that this is some kind of bug (in fact it's more a limitation than a bug) in the Windows wrapper. David Anderson told me that the problem should be fixed in the next release of the BOINC client (it seems that the problem is not solvable at the wrapper level and should be solved at a higher level).
You may wonder why there are some much troubles on Evo and not on other projects using the wrapper. In fact some admins have experienced the same problems, but the consequences are generally less important for 2 mains reasons:
- the problem is due to a sudden kill of the app (due to too excessive RAM consumption for instance) and usually this happens not frequently on stable projects.
- other projects use less RAM so that even if there are more than 1 instance running, the client doesn't notice it.
Anyway, I just need the results of the last 10000 WUs then the problems should be fixed, either by the new BOINC client or by new kind of WUs.
Thanks again for you support.
|
|
| Project News May 20, 2011 |
| Currently we are testing Correlizer jobs and Blender jobs (for the Bridge). Sometimes there are some runs with lots of errors. Of course not what we want, but this happens in large scale tessting. Thanks for helping us with this . |
|
| We are back !!! :) |
|
I have great news for you we are back. We didn't get our server back yet, but we have HDD with all backups.
Now I'm restoring project from backup I think it will take about 1 day so we should start sending wu's tomorrow afternoon. Backup is from 4th Aprli 2010 23:30 CEST so there will be some point loss :( (Server was turned off about 16:30 CEST 5t Aprli)
Sesef
|
|
| Project News May 5, 2011 |
| The new version of BOINC has been installed. We will start with some test jobs. |
|
| CloudSigma does not extend our sponsoring agreement |
During the last months the Freehal Project hosted some free servers at CloudSigma, a cloud computing provider from Switzerland.
Unfortunately CloudSigma decided not to extend our sponsoring agreement. But nevertheless we plan to distribute new workunits in June.
It would be very helpful if somebody who has some bandwidth left could provide a http mirror for us. Please contact us at info@freehal.org if you do.
Maybe someone also would like to sponsor a server for example for a month, as members from SETI.Germany and Planet3DNow! did.
We wish CloudSigma much luck and success in the future.
--
Best Regards,
Tobias Schulz |
|
| Project News April 27, 2011 |
| From tomorrow (April 28) the server will go offline for server maintanence. The services will be unavailable for at least two days. |
|
| Project News April 10th, 2011 |
| To decrease the size of the database, I have manually deleted all the results older than 10 days. Every user should be able to merge unnecessary host entries like double hosts or ghosts.
The problem with the validator is still not solved, he needs sometimes a little kick. ;) |
|
