BOF 2: (BATCH SCHEDULING)
        use BOINC's reliable mechanism to improve turn around on important workunits
                set server parameters to escalate priority of additional results generated for a workunit, these get sent to clients which return work quickly and reliably

        tell the hosts to communicate with server every X days
        if a host doesn't communicate with the server within that time frame (maybe + Y), good chance it's gone 

        how can we tell that a computer will be around?
                keep track of uptime
                many are not around on weekends