• Overview
• Windows *
• Linux *
• Mac *

Using SSH

Users who have installed an HPCMP Kerberos Client Kit and who have a Kerberos ticket may then access many systems via a simple Kerberized ssh, as follows:

%ssh user@system

For some systems, however, you may have to specify a numbered login node. Please review the table below, to get specific system login information.

Information about installing Kerberos clients on your Windows desktop can be found in the Kerberos & Authentication section of this page.

Video Tutorial

A video tutorial is available on logging into a system *.

Using HPC Portal

Information about the HPC Portal may be found on the HPC Portal page.

The HPCMP Centers Team provides an assortment of classified and unclassified computational, storage, visualization, and support resources for DoD scientists and engineers. Please select the Systems tab in the main menu bar to find detailed information about the equipment we make available to users.

While the specific computing environment on our HPC systems may vary by vendor, architecture, and the DoD Supercomputing Resource Center (DSRC) at which the systems are located, we provide certain common elements to help create a similar user experience as you move from system to system within our Program. These elements include environment variables, modules, math libraries, performance and profiling tools, high productivity languages, and others.

Each HPC system consists broadly of a set of login nodes, compute nodes, home directory space, and working directory (scratch) space, along with a large suite of software tools and applications. Access to HPC systems is typically gained through the use of a command line within a secure shell (ssh) instance. Specific authentication and login steps are provided in the Kerberos & Authentication section of this page.

Each DSRC operates a similar petascale mass storage system for long-term data storage. We also provide short-term storage on HPC systems themselves and on a Center-wide File System (CWFS) located at each DSRC.

Need to compile your own source code instead of using the COTS and GOTS applications available on our HPC systems? No problem. Each HPC system offers multiple compiler choices for users. Available compilers and instructions for using them are provided in the "Available Compilers" section of each HPC system's User Guide. The User Guide for a particular system is located on the Systems page; just click the Systems link in the main menu bar above, then navigate to the system of interest and look for the User Guide in the Available Documentation box.

In order to manage the volume of work demanded of HPCMP supercomputers, the HPC Centers Team employs several batch queuing systems for workload management. The batch systems make use of queues, which hold a job until sufficient resources are available to execute the job. The characteristics of the queues found on each HPC system may vary, depending upon the size of the system, the type of workload for which it is optimized, the size of the job, and the priority of the work. To see details of the queues on specific HPC Systems, select the system of interest from the Systems menu in the main menu bar. Look for the "Queue Descriptions and Limits" box.

In a typical workflow, a user submits a job to a queue, and then at a future time when resources are available, a scheduler dispatches the job for execution. Upon completion, the job ends and relevant files are collected and deposited in a location specified by the user. The user generally has no control over when the job starts. If such control is needed, the HPC Centers Team provides the Advance Reservation Service (ARS), which allows users to choose a future time at which the job is guaranteed to run. Note, however, that the number of CPUs dedicated to ARS is limited.

The priority assigned to each queue is dictated by the priority of the work the queue is allowed to run. DoD Service/Agency computational projects may have different types of accounts and may run at different priorities. All foreground and background usage will be tracked by project and subproject and reported to pIE by subproject.

To manage the volume of work demanded of HPCMP supercomputers, the HPC Centers Team employs a batch queuing system for workload management. To run a job in a typical workflow, a user submits a job to a queue, and then at a future time when resources are available, a scheduler dispatches the job for execution. Upon completion, the job ends and relevant files are collected and deposited in a location specified by the user. The user generally has no control over when the job starts. If such control is needed, the HPC Centers Team provides the Advance Reservation Service (ARS), which allows users to choose a future time at which the job is guaranteed to run. Note, however, that the number of CPUs dedicated to ARS is limited.

Batch jobs are controlled by scripts written by the user and submitted to the batch queuing system that manages the compute resource and schedules the job to run based on a set of policies. Batch scripts consist of two parts:

1) a set of directives that describe your resource requirements (time, number of processors, etc.) and
2) Linux commands that perform your computations.

These Linux command may create directories, transfer files, etc.; anything you can type at a Linux shell prompt.

Please refer to each HPC system's queuing system guide for details regarding the format and execution of batch scripts. The queuing system guide for a particular system is located on the Systems page; just click the Systems link in the main menu bar above, then navigate to the system of interest and look for the queuing system guide in the Available Documentation box.