摘要
气象预报是高性能计算的重要应用领域，要用处理能力很强的并行计算机系统才能实现。对于地市级气象部门来说，由于资金的限制，使用超级并行计算机进行气象计算是很不实际的。本文将并行机群技术与目前国际上流行的中尺度气象模型MM5相结合，实现了基于机群系统的中尺度大气数值模型计算系统，为进行中小尺度气象计算的地市气象部门提供了一种经济、可靠的计算方式。
Abstract: Existing Chinese-made weather forecast systems, though cheaper than imported ones, are still too expensive for a developing country like P. R. China. We propose a high-resolution weather forecast system that retains high-resolution but makes the cost of the system only about 50% of existing Chinese-made systems. We hit upon our proposal after having been pressed by users to reduce cost further without lowering the requirement on performance. During the discussion of the users' requirement we found that enough progress in both PC and computer cluster technology has made it possible to satisfy the users' tough requirement on performance-cost ratio. Section 3 discusses how to implement our proposal in great detail; in particular, we connect the parallel cluster PCs with the popular mesoscale model MM5 and implement the mesoscale model numerical computing. Results of testing an eight-node system designed by us are listed in Table 2. Since the beginning of 2001, our system has worked smoothly and steadily in a meteorological station in Huzhou, Zhejiang, P.R.China.
出处
《西北工业大学学报》
EI
CAS
CSCD
北大核心
2002年第1期104-108,共5页
Journal of Northwestern Polytechnical University
