Vikram-100 has a 300 TB of high performance GPFS (General Parallel File System) based storage. The GPFS architecture is good at processing parallel I/O from many nodes in general. However, it is slow when different nodes try to write to exactly the same area of the same file.

The general rule is to avoid having hundreds or thousands of tasks trying to modify the same file/directory at the same time with certain operations. This happens for instance when, on job start, all participating nodes try to create a file each in one and the same directory. A directory is nothing but a file as well.

It is strongly recommended to not do this for any larger job. As as better alternative, the files for the individual tasks can be created all by one task. If the nodes need to create their files indeed themselves, then do create subdirectories first, either one for each tasks or one for a (small) subset of tasks, and let then the tasks create their files within these subdirectories. The subdirectory creation should again be done just by one task.

The code using MPI should do something like this pseudo-code:

!# serial creation
barrier
if (task==0) then
do i=0,nprocs-1
create subdirectory(i) 
create file(i) // with optional truncate option
enddo
endif

!# all files created now
barrier

!#parallel usage
open file(myid)
open file(commonfile_id)
...
write privatefile
write commonfile

Backup Policy

Your data stored in HPC is backed up daily. The data is backed up to a LTO-6 robotic tape library with 48 slots and 4 drives. Depending on data size and number of tapes available, your data backup is stored for at-least 2 months, within which you may request for a retrieval.

Please note that backup process starts at mid-night every day. If you lose your file within few hours, we may not be able to retrieve it.

Storage Quota

All HPC users currently have a storage quota of 3 TB. Previous | Next