Joe Segur made a relevant observation on the SETI message board. When the 'whole system' conveyor belt of jobs is interrupted, all components come to a halt. New work can't enter the system until old work is fully processed and the disk storage space recyled.
Judging by Scarecrow's graphs, it looks as if, for SETI, the current disk storage space is equivalent to a server cache of approximately 48 hours - much lower than many client caches. After that time, there is no point in requesting new work for the stalled application, because there won't be any. That doesn't negate my point that jobs for alternative applications under the same overall project might still be running smoothly on their own distinct conveyor belt. >> I also think taking a solely client-centric view is a mistake. BOINC is >> a system, and you can't optimize the client to the detriment of the >> other components -- you have to optimize the system as a whole. > Agree. I consider proposed change as whole system optimization. _______________________________________________ boinc_dev mailing list boinc_dev@ssl.berkeley.edu http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev To unsubscribe, visit the above URL and (near bottom of page) enter your email address.
