The ARC cluster

From Italian ARC
Revision as of 10:07, 11 April 2014 by Massardi (talk | contribs) (Created page with "The ARC is equipped with a dedicated computer server connected with a high-speed optical fiber network to the outside world, allowing fast data transfer (1 Gbit/sec in 2009; 1...")
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

The ARC is equipped with a dedicated computer server connected with a high-speed optical fiber network to the outside world, allowing fast data transfer (1 Gbit/sec in 2009; 10 Gbit/sec by 2012). Presently we have 16 Tb of diskspace and one 12-unit blade cluster (96 cores) dedicated to the ARC.

ALMA and CASA users can request to access the server and the disk space by sending an e-mail to help-desk@alma.inaf.it indicating the reason for your request.


User policy

ARC users can access the Italian ARC node computing facilities by requesting a face-2-face visit (ALMA users only, through the ALMA Helpdesk) or by visiting the ARC node in Bologna (for any data-reduction-related issue to be solved in collaboration with the ARC staff). In both the cases they are requested to send an e-mail to help-desk@alma.inaf.it indicating the reason for the request.

Please notice that the request for a new account implies that the requesting user (and/or his/her collaborators) visit the ARC for an induction on the ARC facilities usage and on issues related to data reduction with CASA both for ALMA and any other telescope. If the request is positively evaluated the visit details will be arranged via e-mail.

The account will guarantee the usage of the facilities and the support for 6 months. Every user can use 1 TB of disk space. Once the account expires ALL DATA WILL BE REMOVED. Extensions of the account duration period could be considered on request (via e-mail sent in advance with respect to the account expiring date). No visit is needed in case of account renewal.

The ARC members support is guaranteed for any ALMA-related issue. For data-reduction-related issues that do not involve ALMA, the support (other than the technical support in the usage of the ARC computing facilities) is limited to the knowledge/experience and availability of the ARC members.

Upon explicit request IRA staff people can have an account with unlimited duration. IRA collaborators with temporary positions can have an account for the entire duration of their position.

To ensure a well-balanced load on the cluster nodes please follow instructions about accessing the computer cluster

Accessing the computer cluster

Once you have obtained an ARC account at IRA you can access the computer cluster nodes from everywhere through scheduler.ira.inaf.it.

Using graphical applications on the cluster is possible through remote X access. The working nodes are arcbl01 ... arcbl12. Never submit workloads to arcserv (control node) and arcnas1/arcnas2 (storage) as this can slow down the entire cluster.

Access is done via a Torque/Maui scheduler that redirects your job on the less-loaded node.

Jobs on the cluster are limited to a duration of 72 hours.

Here you can monitor the status of Torque scheduler and the number of active jobs. Here you find some statistics about resources consumption on the arcblXX nodes.


Executing programs

You can execute programs in three ways:

  • interactive mode - your command is immediately executed on the less-loaded node and standard input, output and error are linked to your terminal. You can enter a node for interactive work by typing

ssh -tX scheduler.ira.inaf.it

TIP: by typing 'hostname' you can know on wich node you are
   Scheduling a pbs job - submit a job file (here is a guide)
   Scheduling a grid job - you can also use the arc cluster through the grid infrastructure. Just add in you JDL job script the requirement:
 Requirements =  ( RegExp("ira.inaf.it",other.GlueCEUniqueId));


Copying files to from the ARC storage

You can put and get files from outside with scp on the cluster via storage.alma.inaf.it:

scp user@storage.alma.inaf.it:/remote/path /local/path and

scp /local/path user@storage.alma.inaf.it:/remote/path


Mounting ARC storage on you workstation

ARC filesystems can be seamlessly accessed on your workstation with fuse-sshfs:

as root, install the package
(rpm) yum install fuse-sshfs
(deb) apt-get install sshfs
then, as user
sshfs storage.alma.inaf.it:/remote/path /your/local/mount/point/

By omitting /remote/path you can mount you home directory. i.e.:

sshfs storage.alma.inaf.it: /your/local/mount/point/

Be aware that this method is suboptimal for heavy input/output loads. Running disk-intensive applications directly on the arc cluster will result in a file access speed 10-50 times faster.