Client-Server Software
A Key to Success: Client-Server Mode
When a researcher generates data on a modern HPC platform, the amount of data produced can be massive - ranging from tens of gigabytes to hundreds of terabytes. Developing strategies to visualize such data is challenging. Critical decisions must be made early into the project, such as: What visualization/analysis tool should I use? When making a selection, an important question should be kept in mind: Does this software run in client-server mode? Software packages that support client-server mode eliminate the need to move massive amounts of data to a user's local workstation for visualization/analysis. This task would be impractical from a time point of view, if not virtually impossible from a storage point of view. When running software in client-server mode, the client runs on a local workstation while the server runs on an HPC resource where the data was generated. The client and server communicate with one another over the network, so network speed also plays a factor in the usability of such tools. The server performs intensive computations directly on the data, such as, isosurface generation, slicing, ray casting, etc. The client provides the graphical user interface (GUI) allowing the user to interact with the data in the course of normal investigating. The necessary communication between the server and client is handled in a lightweight and intelligent fashion.
Analysis Packages that Go Above and Beyond
Of those packages that do run in client-server mode, three have risen to the top within the HPC realm. There are two reasons for this. Firstly, the servers scale well across hundreds to thousands of compute nodes. Secondly, they provide support for HPC Job Launching, which means they simplify the process of launching the server on a remote HPC resource and establish the necessary secure connection to the client running on a researcher's local workstation. Such job launching hides complex details from the user such as setting up two-way ssh tunnels, Kerberos authentication, and PBS scripts for job submission. The goal is to make running these visualization applications as easy and straightforward as possible. To this end, a simple GUI is presented where basic information, such as number of nodes, number of processors per node, user ID, project number, etc., can be given before initiating the job. The DAAC supports three visualization/analysis software packages that run very well in client-server mode on HPC resources.
These packages are: EnSight, ParaView, and VisIt.
Getting Customized Client-Server Installation Versions
Researchers are welcomed to delve into the details of setting up HPC Job Launching connection scripts for these packages themselves, but this is not necessary. The DAAC has already forged ahead in this area. The DAAC maintains special tarred (and zipped) files, which can be downloaded and extracted anywhere on your local workstation (Linux, Mac, and Windows) - without the need for special admin privileges.
Catalina Users: Seeing a popup similar to this one?
Visit this page for a solution.
EnSight installation tgz/zip files (for Linux, Mac, and Windows workstations) can be installed from the links below. For setup instructions, see the readme.
CRITICAL ANNOUNCEMENTS:
- EnSight no longer allows server-side license checkouts!
Running EnSight now requires setting up a tunnel to the HPCMP’s license server. See instructions here. - EnSight 2022R2 is the last version that supports Mac users!
Default Release
Newest Release
ParaView installation tgz/zip files (for Linux, Mac, and Windows workstations) can be installed from the links below. For setup instructions, see the readme.
Current Release
Newest Release
VisIt installation tgz/zip files (for Linux, Mac, and Windows workstations) can be installed from the links below. For setup instructions, see the readme.
Note: Catalina Users
Seeing a popup similar to this one?
Visit this page for a solution.