作者: Liqiang Cao , Xu Liu , Xiaowen Xu , Zhanjun Liu
DOI: 10.1109/IPDPSW.2019.00079
关键词:
摘要: A log archive for scientific computing applications is a set of logs model and time jobs in HPCs. We have developed light weight fast performance analysis tools on top archives. classify the job based similarity input models to form model-based tree like archive. With linear regression, we analyze relations step with parameters model. found that although there some disturbance, most showed good regularity. In one applications, changes proportionally geometric And significant physical parameter determines up 1.7 times. another application, find each scales 1.59 times number process from 384 768.