数据中心在线指标分析及其仿真计算
On-line index analysis and simulation of data centers
摘要：
可靠性是一种内在属性，是故障次数的概率表达。可用性对故障频率并不敏感，它与运行维护水平有关，可靠性低的数据中心的可用性未必低。可靠性和可用性的平衡实质是故障率与平均故障修复时间的平衡。采用蒙特卡罗仿真模拟方法分别计算了无/有风冷直膨空调的数据中心的可靠性：前者的可用性为99.99%，故障率为1.06×10-4次/h，可靠性（8 760 h）为0.4；后者的可用性为99.86%，故障率为3.6×10-4次/h，可靠性（8 760 h）为0.04。蒙特卡罗仿真模拟方法对电宕机和热宕机的可用性或可靠性分析具有明显优势。
Abstract：
Reliability is an inherent property and a probability expression of the number of failures. Availability is not sensitive to failure frequency, but related to the level of operation and maintenance, which means the availability of data centers with low reliability is not necessarily low. The essential of the balance between reliability and availability is the balance between failure rate and average repair time. Calculates the reliability of data centers without and with air-cooled direct-expansion air conditioning by Monte Carlo simulation method. The availability of the former is 99.99%, the failure rate is 1.06×10-4 times/h, and the reliability (8 760 h) is 0.4. The availability of the latter is 99.86%, the failure rate is 3.6×10-4 times/h, and the reliability (8 760 h) is 0.04. Monte Carlo simulation method has the obvious advantages in the availability and reliability analysis of electrical and thermal downtime.
Keywords:datacenter,availability,reliability,MonteCarlomethod,simulation