Sequence-Dependency-Concerned Process Migration Design of Multilevel Data Centers Under District Power Supply Failure

Data centers (DCs) play a crucial role in the modern digital era. High service reliability is critical for DCs to provide the desired quality of service. This work models the task execution progress from a novel perspective of process sequence dependency, which fully considers the execution sequence of all the constituent processes for the completion of a task. Subsequently, an innovative sequence-dependency-concerned process migration mechanism is established, by which the characteristics of sequence-free tasks, sequence-strict tasks, and sequence-relied tasks are meticulously portrayed. Furthermore, taking a district’s single-point and multipoint power supply failures into account, by exploring the roles of power supply systems (PSSs), backup power systems, and uninterruptible PSSs of DCs with different power supply importance levels, a data-density-based dispatch model is proposed under the district power supply failure. Finally, considering prediction errors of photovoltaic power and process arrival rate in practical DC operations, a two-stage robust optimization model is proposed and a nested column-constraint generation algorithm is employed for the problem solutions. Simulation results verify the presented task migration mechanism can effectively conduct economic task reallocation following the task completion priority and sequence dependency requirements.