File Transfers

From HPCC Wiki
Jump to navigation Jump to search

There are several methods for transferring files to HPCC depend on tier and file system used.

Basic and advanced tier file transfer

Basic and advanced tiers are attached to DSMS file system as it described in figure 1. above. The users utilizing these resources may transfer files via Globus online or via Cea - the file transfer node as it is described below In addition the users of these tiers may use iRODS data management system as described below.

Globus Online: The preferred method for large files, with extra features for parallel data streams, auto-tuning and auto-fault recovery. Globus online is to transfer files between systems—between the CUNY HPC Center resources and XSEDE facilities, or even users’ desktops. A typical transfer rate ranges from 100 to 400 Mbps.
Details on connecting via Globus can be found here.
cea.csi.cuny.edu: Only Secure FTP (SFTP) can be used to transfer files to cea.csi.cuny.edu. Each of the servers is mounted on cea under its own name. Thus the users could place their files on particular server simply by preceding /scratch with /<name of the server> e.g. sftp> put TEST.txt /penzias/scratch/john.doe/TEST2.txt
Use sftp to transfer files to/from Penzias file systems
iRODS: The data grid/data management tool provided by CUNY HPC Center for accessing the SR1 resource. iRODS clients (iput, iget, irsync) provide a data transfer mechanism featuring bulk upload and parallel streams.

Transfer files to/from Arrow cluster:

For the first project year the storage purchased with NSF grant will be accessible only by the core members of the NSF grant. For that time the acquired storage system will not be applied as main storage for HPCC and consequently the files can only be tunneled from/to Arrow. Note that files must not be explicitly copied on Chizen and then copied to Arrow. Files copied to Chizen and left there will be deleted.

Users can use tunneled sftp secure shell to tunnel their files from/to /scratch directory on Arrow. To put or get files from local machine to directory in Arrow please open sftp dialog by typing:

sftp -J <user_id>@chizen.csi.cuny.edu  <user_id>@arrow:/scratch/<user_id>

Then provide twice the user's password. e.g.

(<user_id>@chizen.csi.cuny.edu) Password:

(<user_id>@arrow) Password:

Changing to: /scratch/<user_id>

sftp>

The first password will be used to open connection between local server/desktop/laptop and chizen and the second to establish tunneled connection to arrow server. The files can be obtrained from /scratch/<user_id> with command get and put there with command put. For example, the following dialog shows how to get copy of file test.sh from /scratch/<user_id> to local directory and changing the name of the copied file to test5.sh.

sftp> get test.sh test5.sh

Fetching /pfssfs1/scratch/<user_id>/test.sh to test5.sh

test.sh                                                                                             100%  172     8.7KB/s   00:00    

sftp> bye


The users who want only (one directional) quick copy from local machine to arrow:/scratch/<user_id> can also use the tunneled secure copy scp via chizen by typing:

scp -J <user_id>@chizen.csi.cuny.edu <file_to_transfer> <user_id>@arrow:/scratch/<user_id/.

Note that white symbols work for sftp.