基于数据中心气流组织的节点失效模型研究与设计-计算机科学与技术计算机技术专业论文.docxVIP

暨南大学硕士学位论文 暨南大学硕士学位论文 万方数据 万方数据 摘 要 在传统的数据中心中空调制冷系统的能耗占据了总能耗很大的一部分。由于对其不合 理的操作，往往增加了空调制冷系统冷却能耗，造成不必要的能源浪费。另一方面，数据 中心节点失效后，节点执行任务的可靠性无法保证，若不采取有效措施，将会发生不可估 计的损失。 结合以上问题，本文在数据中心气流组织的模型[1]的基础上，建立基于数据中心气流 组织的节点失效模型，并提出基于节点失效模型的任务调度算法。在节点失效模型中，文 章主要证明了节点失效模型中的交叉干扰系数矩阵和负载均衡的限制条件，并介绍了基于 节点失效模型的智能算法。 本文在 Matlab 平台上对文章提出的基于数据中心气流组织的节点失效模型进行模拟 仿真，实验验证了基于数据中心气流组织的节点失效模型的适用性，实验结果表明失效节 点的增多会导致数据中心的能耗增加，数据中心的能耗与失效节点在机架中的位置有关。 关键字： 气流组织；节点失效；任务调度算法 I Abstract In a traditional data center, energy consumption of computer room air conditioning systems occupies a large portion of the total energy consumption. Due to unreasonable operation, the cooling energy consumption of computer room air conditioning systems tend to increase, and unreasonable operation lead to unnecessary waste of energy. On the other hand, after the data center’s nodes failure, reliability of nodes failure can’t be guaranteed. If we fail to take effective measures, it will occur serious losses. Combining these issues, we build a nodes failure model which is based on data center’s airflow distribution, we also propose a task scheduling algorithm about the nodes failure model. In the nodes failure model, the paper mainly proves cross-interference coefficient matrix, constraints based on load balance and introduces intelligent algorithm based on the nodes failure model. We simulate the nodes failure model in Matlab environment. According to experimental results, we verify the applicability of the nodes failure model, Experimental results show that increasing the failed nodes will lead to increasing energy consumption of data center which is associated with failure nodes in racks location. Keywords: airflow distribution; nodes failure; task scheduling algorithm II 目 录 HYPERLINK \l _bookmark0 摘 要 ......................................................................................................................... I HYPERLINK \l _bookmark1 Abstract ............................................................