[mvapich-discuss] mvapich2 error/romio/PMB-IO
TaoJinZhong
com91tao at hotmail.com
Tue Oct 16 06:33:05 EDT 2007
Hello all: I configure,build and install mvapich2-1.0 on the Linux cluster¡£The system configurations are listed in the following: 2.6.9-42.ELsmp/4*Intel(R) Xeon(R) CPU/16G Memory/Lustre filesystem The hca driver is based on OFED-1.1.1.I use make.mvapich2.ofa to install the mvapich2-1.0,and modify the configure line to include "--with-romio --with-file-sysmtem=nfs+ufs" to support ROMIO. When i run the PMB-IO(Pallas MPI Benchmarks 2.2.1),the erros are following:_________ [root at cn001 SRC]# /home/mvapich2-io/bin/mpiexec -n 4 ./PMB-IO #----------------------------------------------------------------------------- # Benchmarking P_IWrite_Indv # ( #processes = 1 ) # ( 3 additional processes waiting in MPI_Barrier) #----------------------------------------------------------------------------- # # MODE: AGGREGATE # #bytes #repetitions t_ovrl[usec] t_pure[usec] t_CPU[usec] overlap[%] 0 50 94918.84 36.32 89349.03
0.00 rank 0 in job 39 cn001_38658 caused collective abort of all ranks exit status of rank 0: killed by signal 9 --------- 1.Because the ROMIO module don't support the Lustre Filesystem,i want use the UFS(romio) to instead.The execution PMB-IO is in the lustre filesystem.And the partial results are also seem illogical,for example:-------- #----------------------------------------------------------------------------- # Benchmarking P_Write_Indv # ( #processes = 4 ) #----------------------------------------------------------------------------- # # MODE: AGGREGATE # #bytes #repetitions t_min[usec] t_max[usec] t_avg[usec] Mbytes/sec 0 50 22.30 25.66 23.61 0.00 1 50 1.66 8.96 3.67 0.11 2 50 1.44 8.18 3.28 0.23 4 50 1.56 8.66 3.47 0.44
8 50 1.52 10573.14 5252.95 0.00 16 50 16597.92 16853.08 16745.10 0.00 32 50 16120.32 16398.58 16276.92 0.00 64 50 17287.30 17557.16 17433.27 0.00 128 50 16806.64 17497.32 17267.02 0.01 256 50 5410.92 16623.88 13716.83 0.01 #------------------------------------------------------------------------------------- Is it because of the use of UFS(romio) in the lustre filesystem? 2.what is the reason of the erros:"rank 0 in job 39 cn001_38658 caused collective abort of all ranks"? 3.Has anybody run the PMB-IO(PMB-2.2.1) test with the mvapich2? taojzh
²»µÇ¼¾ÍÄܹÜÀí¶à¸öÓʼþÕÊ»§£¬ÊÔÊÔ Windows Live Mail¡£ Á¢¼´³¢ÊÔ£¡
_________________________________________________________________
Óà Live Search ËѾ¡ÌìÏÂ×ÊѶ£¡
http://www.live.com/?searchOnly=true
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.cse.ohio-state.edu/pipermail/mvapich-discuss/attachments/20071016/8c47694c/attachment-0001.html
More information about the mvapich-discuss
mailing list