摘要
分析了以超级计算中心联想深腾6800为主的HPCS监控需求,比较了大量的监控实现技术,给出监控系统评估特征和指标,提出了一种集系统、性能、应用程序、进程监控于一体的改进策略,讨论了利用信息流水、过滤、双重传输模式减少监控数据传输量,减轻监控资源消耗,提高综合监控的性能和效率。
The monitoring requirement on Deepcomp 6800 and other HPCS was analyzed, compares with a large number of existed monitoring techniques, presents the evaluation features and criteria of monitoring system, then advances an improved strategy that integrates system, performance, application and process monitoring together, finally discusses using pipelining, filtering and double-mode transfer to reduce monitoring data, alleviate monitoring overhead, enhance monitoring performance and efficiency.
出处
《计算机应用研究》
CSCD
北大核心
2007年第7期24-27,共4页
Application Research of Computers
基金
国家自然科学基金资助项目(60533020)
国家"973"计划资助项目(2005-CB321702)
关键词
HPCS
监控
Clumon+
过滤
双重传输模式
high performance computing systems(HPCS)
monitoring
Clumon +
filtering
double-mode transfer