دانلود رایگان مقاله شبکه مرکز داده شکست آگاه منطقه ای احتمالاتی و قرار دادن محتوا

Abstract

Data center network (DCN) and content placement with the consideration of potential large-scale region failure is critical to minimize the DCN loss and disruptions under such catastrophic scenario. This paper considers the optimal placement of DCN and content for DCN failure probability minimization against a region failure. Given a network for DCN placement, a general probabilistic region failure model is adopted to capture the key features of a region failure and to determine the failure probability of a node/link in the network under the region failure. We then propose a general grid partition-based scheme to flexibly define the global nonuniform distribution of potential region failure in terms of its occurring probability and intensity. Such grid partition scheme also helps us to evaluate the vulnerability of a given network under a region failure and thus to create a “vulnerability map” for DCN and content placement in the network. With the help of the “vulnerability map”, we further develop an integer linear program (ILP)-based theoretical framework to identify the optimal placement of DCN and content, which leads to the minimum DCN failure probability against a region failure. A heuristic is also suggested to make the overall placement problem more scalable for large-scale networks. Finally, an example and extensive numerical results are provided to illustrate the proposed DCN and content placement.

6. Conclusions

We studied the DCN and content placement problem under global nonuniform distribution of potential region failure due to disaster in large-scale geographical areas. By proposing a general grid partition-based vulnerability estimation scheme, we can determine the “vulnerability map” of a given network for DCN and content placement, which provides an important input for our proposed ILP and heuristic. Based on the vulnerability map, our proposed ILP can generate optimal DCN and content placement solutions to minimize the DCN failure risk due to disaster. This achieves best-effort protection of DCN and content against the region failure. To make our solution more scalable for largescale networks, a heuristic was further proposed. Numerical results showed that our work can lead to a more feasible solution. It can well protect DCN and content under global nonuniform distribution of the potential region failure scenario.