内容分块数据中心光网络灾前数据转移

Data Evacuation in Content-Fragmentation based Optical Cloud Networks before a Larger-Scale Disaster

大范围灾难的发生将导致多个数据中心同时毁坏,现有内容备份策略无法完全恢复所有数据。因此,在灾难发生前需对灾难区域内数据进行转移。文章研究了内容分块(CF)数据中心光网络灾前的快速数据转移问题,构建了混合整数线性规划(MILP)模型并提出了最小时延(LD)快速数据转移算法(简称LD算法)。LD算法的性能与MILP模型非常接近,整个网络数据转移时间接近最优解。此外,仿真结果表明,在相同数据冗余度下,采用LD算法的CF数据中心光网络比内容复制数据中心光网络最多节省42%的数据转移时间。

A large-scale disaster will affect multiple data centers of an optical cloud network and the contents stored in these data centers should be evacuated for recovery as soon as possible. This paper addresses the problem of data evacuation in a Content-Fragmentation(CF) based optical cloud network before a large-scale disaster. We formulate this problem as a Mixed Integer Linear Programming(MILP) model and develop an efficient Least Delay(LD) based evacuation algorithm. Numerical results show that the performance of proposed LD algorithm is similar to that of the MILP model. In addition, compared to the content replication scheme, the CF based optical cloud network using LD algorithm always takes shorter times to evacuate data.