How to reduce the number of “lost� WU’s

Don’t think i have seen any thing about what the project does to reduce this in the forums, so hear goes.

Then the client requests more work from the server, the server checks if the host is likely to finish the requested work before deadline. As part of this i believe the server asks the client how many Einstein WU’s that is already stored on the computer, so it can calculate if the additional requested work is likely to be done before the deadline.

1. My first suggestion is: then the client requests more work, the server first compare how many Einstein WU’s (including unsent results) that is stored on the computer, with how many the database sees it should be. If these numbers don’t mach, the server should compare what WU’s that actually are on the computer against the database. WU’s not found on the computer would then be marked with an error massage so a replacement WU can be sent out to a new host.

This would reduce the number of “ghost� WU’s and WU’s lost then resetting the project. It would also detect WU’s deleted bye the users of 4.2x clients.

2. Then resetting/detaching from the project, the scheduler should be informed that the client has done this so it can recycle the “lost� WU’s.

3. The first time BOINC is installed on a computer a random computer identification number should be generated. This would help detecting if a computer have trouble attaching to the project and there by creating multiple hosts on his account for the same computer. I think I saw a user who had a computer that immediately detached every time he tried to attach to the project. Every time he tried, a WU was recorded as downloaded by the new host.