Availability Evaluation of Dawning Cluster Job Management System
Nie Hua · Jisuanji fangzhen · 2007
To meet the new requirement of continuous development of HPC Cluster,we design Dawning Cluster Job Management System-DCJM,and implement several layers of high-availability mechanism. To evaluate its availability,we integrate RBDM,Fault Tree and Stochastic Petri Net to build a comprehensive model for DCJM. Then we resolve the model and analyze computation value of the system model. The numerical evaluation result shows that DCJM has achieved a high availability.