I have been thinking about what the appropriate automatic sizes for work fetch 
and EDF trigger are for work fetch and for EDF for rr_sim.

If we track the longest duration (over the past N months) from the first time 
that a task reports as 100% complete till report is complete pre project, then:

The minimum required work buffer is the smallest of these across all projects 
able to fetch work.  Eg.  Project A has a 1 week value, project B has a 1 hour 
value, the minimum work buffer would be 1 hour.

For each project, any task that exceeds report deadline for the task - the 
longest durataion from 100% to report complete needs to be marked as requiring 
EDF.

It would also be useful if the projects could have a tag that could be set to 
indicate they expected to be offline for N seconds (probably entered as days 
and converted to seconds).  The maximum of this and the largest duration 
already recorded would be used for this.

For example, SETI would normally have this set to 0.4 days to indicate the 
weekly outage.  Occasionally SETI could bump this number to allow clients to 
cope with expected longer outages.  Yes, unexpected outages would not be helped 
by this.
_______________________________________________
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.

Reply via email to