[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [bgl-discuss] Blue Gene slow I/O



Could you give us a few more details? Is each process writing to a separate file, with all processes writing at about the same time? Are they writing contiguous blocks to the same file, with the blocks aligned on block boundaries? Are they writing multiple segments to the same file, with each process contributing some set of segments?

Do you use MPI-IO? If so, which MPI-IO routines do you use? If not, would you be willing to consider using MPI-IO?

Thanks!
Bill

On Dec 4, 2006, at 10:13 AM, Sindhi, Abdul Waris wrote:

We are running a CFD application on the Blue Gene that performs sequential
I/O on all the processors at around the same time. The case we currently run
is out of user homedir. We are currently running a 40 processor breakup, and


the read/writes are to a file sized around ~20MB.

The read/write performance looks pathologically slow at around 20-30 times
slower when compared to normal runs. We are wondering if anyone else has
run into similar issues with slow I/O on the Blue Gene ?


Thanks,
A. Waris Sindhi
High Performance Technical Computing
Pratt & Whitney, UTC
(860)-565-8486

- --------------------------------------------------------------------
To add or remove yourself from this mailing list, use the 'notifyme'
command on any BGL machine. To remove: notifyme -n, to add: notifyme -y.



- -------------------------------------------------------------------- To add or remove yourself from this mailing list, use the 'notifyme' command on any BGL machine. To remove: notifyme -n, to add: notifyme -y.