In the real world computers are often unavailable: they may be connected via a dialup line or suffer from frequent soft failures (e.g., Windows crashes). Partners must agree on a level of required uptime (e.g., ``up 90% of the time'' or ``up during California business hours'').
Lower levels of partner uptime decrease performance: backups and restores take longer because the computer must wait for partners to become available. For example, if a machine's partners are up only during business hours and it crashes during the weekend, no restore will be available until Monday morning. Efficient backups require most partners to be up simultaneously during some period of the day. This limits the ability of computers with low and unpredictable uptime to participate in our scheme.
Agreements are subject to being broken. For the simplified scheme, we assume that owners are not out to take advantage of or hurt others. We do not, however, assume that owners are reliable about maintaining uptime agreements. Owners might forget to leave their computer on as much as planned, underestimate how often their machine crashes, or change their computer-usage policy without remembering to tell the backup-system software.
To guard against this, each computer keeps track of its uptime and warns its owner when it is failing to live up to its end of its agreement. For the simplified scheme, we assume this reminder is sufficient to make the owner take any needed steps to correct the problem. In the full scheme (see Section 3), we actively police agreements (both uptime and storage swapping) and abandon partners who fail to live up to their end of an agreement.