On Sep 28, 2009, at 12:54 PM, [email protected] wrote: > We need to reward only for successful tasks. For computation > resources, it > needs to be based on the FLOP count. This can be estimated in several > ways, and no matter what is done gold plating benchmarks, it will > always be > an estimate.
The issue as originally raised was that there are more and more projects that are looking at or now using BOINC to do things that were originally envisioned but we have never taken the time to specify. Results that are tied to computing can be tied to FLOPs or IOPs but we have never specified how we intend to measure and reward I/O Ops ... or Gigabytes ... or any of the other general resources. There are three NCI project right now that I am aware of FreeHAL, QCN, and Anansi what standard should they use? As to the standard of only successful tasks because that catches cheaters ... can you show evidence that cheaters are being caught this way? Or is it only the poor participant trying to do good work that has his effort invalidated because the O/C club has invalidated all the other results? Since I don't like having the Resource Scheduler run once a second should I now refer to your defense of that mode of operation is such a tone? This is beneath you John. Were I advocating to run the current benchmark for hours to only improve its "accuracy" or to run one that is more elaborate to do that one thing... you might have a point. But, the proposal is much broader than that. Lynn took me out to the wood shed because he feels my tone is wrong ... I have to ask myself why he did not do the same with you? _______________________________________________ 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.
