Dear volunteers,
We are analyzing the first batch of simulations and so far the results are quite satisfactory. 98.35% of the tasks launched once the problems with the validator were solved, run without problems and their results were valid. 1.12% of them failed to boot into the client (all under the same user and host). 0.53% were tasks that were executed satisfactorily but whose results were not validated when comparing whit the results of the other volunteers. We are studying what could have happened in these tasks so that the result did not coincide with those of the other. As soon as everything is checked, we will repeat the test and if everything goes well, we will move on to a larger test.
Thank you very much to all, we keep you informed of the progress.

Best,
Jesus.

===============================================

Estimados voluntarios,
Estamos analizando la primera tanda de simulaciones y, hasta ahora, los resultados son bastante satisfactorios. De las tareas lanzadas una vez solucionados los problemas con el validador, el 98.35% de las tareas lo hicieron sin problemas y sus resultados fueron válidos. El 1.12% de las dieron error al arrancar en el cliente (todas en el mismo usuario). El 0.53% fueron tareas que se ejecutaron satisfactoriamente pero cuyos resultados no fueron validados al comparar. Estamos estudiando qué ha podido pasar en esas tareas para que el resultado no coincidiera con los de los otros voluntarios. En cuanto todo esté revisado repetiremos la prueba y si todo va bien, pasaremos a una prueba de mayor tamaño.
Muchas gracias a todos, os mantenemos informados de los progresos.

Un saludo,
Jesús.

---

Jesús Carro
Universidad San Jorge
@InSilicoHeart

0.53% were tasks that were executed satisfactorily but whose results were not validated when comparing whit the results of the other volunteers. We are studying what could have happened in these tasks so that the result did not coincide with those of the other. As soon as everything is checked, we will repeat the test and if everything goes well, we will move on to a larger test.

It is probably due to overclocking. A lot of users think that because they can run a CPU without crashing, that it is working OK for scientific purposes.

I noticed one of my clients received 256 of the newest beta v0.07 work units (~15%?) due to a misconfiguration of my end (had entered 10 days of work instead of 0.1). While it'll complete the tasks well before the deadline (7 days), I'd have thought it'd be better to have these be spread across a wider range of platforms/ clients. Let me know if you'd want me to abort these tasks (although, it may just go and sit in someone's else queue).

That's good thinking Anand. Usually the idea of a test run is to spread the wu's across a wide range of different computers and OS's to identify problematic combinations. Furthermore if there are long work queues on some machines it also stretches the test run unnecessarily. Presumably Jesús can only continue with the next test when all the results have returned and got validated. I keep my queues at 0.1 or 0.2 days to make sure the wu's are returned speedily.

Hola Jesús! During testing could you maybe limit the numbers of work units every host can receive at one time so that they are returned faster?

Drago75 wrote:

Hola Jesús! During testing could you maybe limit the numbers of work units every host can receive at one time so that they are returned faster?

+1

Hi AnandBhat,
Thank you for cancelling the tasks. Yes, to test the application it is better to have as much computers as possible. I will review the limits in the number of tasks for future batches of simulations.

Best,
Jesús

---

Jesús Carro
Universidad San Jorge
@InSilicoHeart