From: Tiantian Liu (tiantianl_at_gmail_dot_com)
Date: Tue Oct 27 2009 - 17:58:13 PDT
Hi: I have used BLCR for 2 years. And I have achieved "Process Migration" in LAM/MPI Parallel Computing. Now, I do some experiments about my process migration system in cluster. I selected 2 computers(Intel dual-core and 2GB memory) to build a mini-cluster and HPL benchmark to test my system. When I choose the HPL problem size as 2000 and 2 processes running, the BLCR spent a lot of time in doing checkpoint. About 20 seconds the BLCR costed to generate 2 checkpoints for HPL.!! I think it's to long!! I want to know if the BLCR really spends a lot of time when checkpoint large program? Dose anyone do some performance test about BLCR? Woud you mind to give me some test result of BLCR to me for comparison? Thank You!!