Why not add a feature where BOINC can create scenario files from the actual work run on systems.
Thus, while running BOINC those of us that are trying to help improve the system could log the real world scenarios we see and then add those to the library for use and or alteration? John's other point is that comprehensive modeling would in fact require running simulations over long periods to capture the behavior of the system with tasks like CPDN or Orbit where the tasks run for 200-600 hours. Sometimes the problem is that these tasks bias the system in unfavorable ways. On Jan 8, 2010, at 10:43 AM, David Anderson wrote: > [email protected] wrote: >> The last time I looked, there were only a handful of simulations available. >> This is not really enough to see any problems. > > "Available"? The idea is that people create simulation scenarios (XML files) > representing real or hypothetical cases, > and that we build up a library of these. > >> >> A good simulation takes several hours to run. > > Last time I checked, a simulation of 24 hours real time > takes << 1 sec. of CPU time. > _______________________________________________ > boinc_dev mailing list > [email protected] > http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev > To unsubscribe, visit the above URL and > (near bottom of page) enter your email address. _______________________________________________ boinc_dev mailing list [email protected] http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev To unsubscribe, visit the above URL and (near bottom of page) enter your email address.
