TeraStat is organized as a batch system. The user interested in running an experiment has to prepare a script file containing all the instructions of the experiment. This script will be then submitted to the TeraStat queuing system that, in turns, will run it as soon as enough computing resources will be available. 

Support requests can be sent to supercalcolo.dssATuniroma1.it

Connecting to TeraStat

It is possible to access to TeraStat by using a login node. This can be used to establish an interactive working session useful to prepare the scripts describing the experiments to run and for file transfer purposes. There are two possible types of connection.

Terminal-based connection

A terminal-based connection to TeraStat can be established via an SSH client application. This application is pre-installed on Linux and MacOS based systems (i.e., the ssh command). If using a Windows-based system, a third-party SSH client application should be used like, e.g., Putty (http://the.earth.li/~sgtatham/putty/latest/x86/putty.exe).
SCP and SFTP protocols can be used to trasfer files to and from the cluster.

Connections parameters:
hostname: n2.sta.uniroma1.it
port: 22

GUI-based connection

A GUI-based connection to TeraStat can be established via a client application supporting the NX protocol, like NoMachine (https://www.nomachine.com/download) or OpenNX (http://opennx.net/download.html).

Connections parameters:
hostname: n2.sta.uniroma1.it
port: 22
protocol: SSH

Once connected, the user has two home directories to use. The default home directory is located at the path  /afs/sta.uniroma1.it/user/home/ and uses the AFS file system. Moreover, each user has a second home directory located at the path  /lustre/work. This second directory is hosted on a Lustre file system and, thus, guarantees much better access times. For this reason, the user is advised to use this second directory for his data and for his experiments.

The Queue system

The execution of an experiment on TeraStat is done through a queue system. The user submits to this system a script files containing the instructions for the experiment, by using the qsub command. The queue system accept in input this script file and schedules it for execution as soon as the required resources are available on the cluster. The execution order of the submitted scripts is chosen according to a priority-based algorithm and to the particular queue chosen by the user when submitting the experiment. For instances, experiments submitted to the S (small) queue have a higher priority than experiments submitted to the M (medium) queue. The order is also influenced by other factors. For example, a user that has been using the cluster for a long time is assigned with lower priority than a user that has never used the cluster.

Queue Maximum number of nodes Minimum number of cores Maximum number of cores Maximum Duration
S 4 1 128 8
M 1 1 32 48
L 2 1 64 48
XL 4 65 128 48
XXL 4 1 128 99

Module and Software Selection

Different software can be used on the TeraStat cluster. Some are directly available when you log in, while for others you need to configure your shell environment. The module system simplifies this operation.

Four shell commands are available:

  • module list: shows the loaded modules
  • module avail: shows the available modules
  • module load <module name>: loads a new module
  • module rm <module name>: removes a module previously loaded (NB the removal of a module can compromise the operation of other modules due to shared dependencies)

For all these commands the shell autocompletion is active.

The effect of these commands is limited to the shell in which they are executed.

Preparing and submitting an experiment

See https://sys2a.sta.uniroma1.it/torque.htm