Main Page

Introduction

The ARROW cluster at TRACC is being operated for the Federal Highway Administration (FHWA) and the National Highway Traffic Safety Administration (NTHSA) at the US Department of Transportation. The site operates at Argonne National Laboratory under the Transportation Research and Analysis Computing Center (TRACC). The site is funded by specific US Department of Transportation projects and is not available to the general research community.

TRACC has combined the original hardware from the Phoenix and Zephyr clusters into a single HPC cluster called ARROW. This consolidation allows for the efficient administration of TRACC cluster services. To mitigate the problems with load balancing, the different types of hardware nodes on the ARROW cluster are partitioned into queues with specific characteristics. When new hardware is installed to expand cluster resources, it will be made typically available via a new queue.

ARROW is arranged such that there is a set of four login nodes that are accessed by users using the SSH protocol. The login servers are configured identically so that it doesn't matter which login server a user is connected to (the caveat is that the user should connect to the same login server where he may have created a GUI desktop session). All servers and nodes share a single parallel file system, and an individual home directory is provided for each user to store their data.

Some common software packages provided on the cluster are commercial engineering applications such as StarCCM+, and LS-Dyna. In addition, selected open source software is available, such as OpenFOAM. Specific versions of each software are selected by the user through the "module" command, which is commonly used on HPC clusters. Connections are limited to the SSH protocol at this time, and cyber security is a major aspect of our operation. We will work with you to optimize your remote operating experience. Users will be able to create and use GUI desktops using the XFCE software. These desktops will stay active until the users decides to log out. Otherwise, GUI sessions will stay active and can be reconnected to, similar to the Windows Remote Desktop software. Thus, users can connect and disconnect from these sessions at will, and network interruptions will typically not interfere with your work flow. The PBS job submission framework allows users to use HPC resources in an efficient manner, sharing a fairly complex selection of specialty systems. This includes typical HPC nodes, as well as selected GPU systems for AI/ML applications.

The links and pages shown below are outdated

We are updating the wiki pages with the most recent information on services, hardware, and capabilities. The information provided below is outdated and in urgent need for a update. We recommend to disregard the information for now.

• How to Connect
  
• Transferring Files
  
• Setting Up Your Environment
  
• Login Nodes Versus Compute Nodes
  
• Files & Storage
  
• Backups
  
• Compiling Your Code
  
• Running Your Code
  
• Graphical Applications
  

TRACC Cluster Software

• Supported On ARROW
  
• Under Investigation
  

TRACC Policies

• Application Restrictions
  
• Data Retention
  
• Email
  
• Login Nodes
  

---

If you are either not familiar with the MediaWiki collaboration platform or have never edited this wiki before, please visit the Help page. In order to edit this wiki you must first log in.