Least Maintenance Batch Scheduling in Cloud Data Center Networks

摘要

Nowadays, to efficiently safeguard the uptime of data centers, system-scale preventive maintenance is necessary. If all servers are under maintenance at the same time, it will completely disrupt the service continuity of all virtual machines that run over servers, thus it necessitates the proposition of an appropriate multi-batch scheme in which during each batch only a partial set of servers are maintained, leaving the remaining servers still on to fully support the virtual machine operations. To enable the cost-saving multi-batch scheme, in this letter, we aim to minimize the total number of batches to implement the system-scale maintenance, meanwhile guaranteeing the system service continuity. We propose a novel heuristic to solve the problem which can well achieve the problem lower bound.