My point exactly -- in your example you've got a buffer in the middle, and you are maximizing flow between the buffer and the validator.
The only difference between that and my suggestion is that you can do exactly the same thing between the upload server and the clients. -- Lynn Gabor Gombas wrote: > On Tue, Jul 14, 2009 at 10:56:58AM -0400, [email protected] wrote: > Yes, but you now control both ends of the network connection between the > upload servers and the server running the validator, meaning you can > tune it to saturate the link without dropping packets, therefore > maximizing the usage of your resources. > > These kind of buffering between the client and the validator is exactly > the mechanism you need to smooth out the upload spikes, and to hide your > internal problems from the clients. > > Gabor > _______________________________________________ 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.
