Advanced Computing in the Age of AI | Tuesday, June 25, 2024

Altair Upgrades Compute Manager Giving Small to Medium Sized Manufacturers an Assist in Using HPC 

<img style="float: left;" src="" alt="" width="95" height="109" />With the release of Compute Manager 11.1, leveraging HPC to take better advantage of digital manufacturing tools like modeling and simulation just got a lot easier.

Small to medium sized manufacturers (SMMs), have just received another helping hand in their quest to take advantage of the tools of high performance computing and digital manufacturing.  

HPC is no longer the exclusive domain of big enterprises like Boeing, John Deere or General Motors. SMMs are also increasingly relying on the technology to lower costs, empower their engineers, designers and analysts, and be more competitive.

The problem is that many of these smaller companies are either not familiar with HPC or are still at a rudimentary stage where they rely on high-end PCs or workstations to run their modeling and simulation software.  Also lacking is a supporting IT infrastructure, in-house expertise, and that biggest stumbling block of all – money.

Many HPC hardware and software vendors are addressing these problems by providing: simplified, web-based interfaces to their offerings; cloud computing capabilities; and support services that range from on-line help to on-site engineers.

Altair is one of the companies attempting to make life easier for its customers.  Their latest offering, recently released, is a new version of Compute Manager, the company’s HPC portal that simplifies the set-up, monitoring and visualization of simulations.

The initial version of Compute Manager was launched earlier this year along with PBS Professional, Altair’s commercial grade HPC workload management system.  Like several Altair flagship products, PBS (Portable Batch System) Professional has been around for a while.  Originally developed for NASA back in 1991, the intellectual rights to the technology were acquired by Altair in 2003 along with the original development team.  This is big iron software. One of its features is a three-fold boost in scheduling speed – less than 30 seconds per cycle for a typical HPC workload up to 80,000 cores. It can support hundreds of thousands of processors ranging from clusters to the largest HPC systems.

Helping the Missing Middle

But what about the SMMs?  Here’s where the latest version of Compute Manager comes into its own.

Comments Venkat Parameshwaran, Director, Enterprise Solutions at Altair, “Many SMMs lack HPC knowledge. Often what we see is that these companies do not have the systems or IT background to administer and manage HPC clusters.  What Compute Manager provides is a web-based application that hides all the complexities around implementation and use. This allows the engineers who are performing the simulations to achieve their goals pretty easily without having to having to know a lot about how to administer clusters or Windows HPC systems.”

The initial release of Compute Manager provided a number of benefits including:
•    Automation of job submissions and set-up of distributed resources
•    Graphic and analytic monitoring of workloads
•    Easy to use web-interface to submit and monitor jobs in PBS Professional
•    Manage simulation jobs with remote data visualization
o    Automated staging of input and output files
•    Diagnosis or remote jobs without downloading huge results files

The latest release makes life even simpler for engineers and designers, especially those in smaller shops who do not have a HPC modeling and simulation background comparable to their counterparts in larger enterprises.

Results Visualization Service

One of the new features of Compute Manager 11.1 is the inclusion of a formal extension that permits the 3D and 2D visualization of jobs while they are running.  

Says Parameshwaran, “Previously you submitted a job and looked at all the ASCII text files in the viewer – you would have to go through pages and pages of log files generated by the solvers.”  

For example, he continues, you might want to look at an X-Y plot that has 1000 data points derived from an eight-gigabyte file.  You are sitting in your office in California, and your cluster is in Michigan, and it’s going to take a year and a day to download the results. With the Results Visualization Service, some clever server-side co-processing of the visual data can turn a cumbersome file that takes forever to download into information that can be viewed while the job is running.

Parameshwaran explains that with the help of the Compute Manager plug-in, both large enterprises and SMMs are able to see 3D animation and view 2D plots in the form of line graphs in their web browsers.  With the Results Visualization Service, users can visualize and manipulate 2D plots and curves generated from job results data.  This includes working with such computer-aided engineering (CAE) apps as finite element analysis (FEA) and computational fluid dynamics (CFD) available in HyperWorks.

The new version of the software also supports the viewing and management of job arrays.  A simple-to-navigate interface allows easy submission and access to jobs and individual sub-jobs with their specific properties and files.

Access control is another new feature providing administrators with fine-grained control of who has access to what portions of the application.  A centralized configuration system for defining users and their groups allows users to share solver profiles and control access to clusters and applications.


Altair has also made life easier for systems administrators, especially in smaller companies, by providing Compute Manager as an out-of-the-box product with minimal data staging and other configuration requirements.  

Parameshwaran also points out that Altair has a global support force of more than 500 engineers available to help customers with the implementation and use of the company’s portfolio of products and services.

Cloud computing figures heavily in Altair’s plans and the company is currently engaged in an experiment called HyperWorks On-Demand.  Many SMMs are already using HyperWorks products such as HyperMesh and HyperView as well as solvers like RADIOSS and OptiStruct.  HyperWorks On-Demand provides customers with an easy way to access these applications running on HPC resources in the cloud.  

Also, licensing problems, often a major stumbling block to SMMs who want to implement HPC solutions, are minimized by the unique pay-for-use, unit-based business model incorporated in PBS Works, Altair’s suite of on-demand computing technologies.

“All our solutions run on the same token system, so a SMM that is already using HyperWorks running on a small cluster with two or three nodes, can use the same tokens to access a powerful HPC system in the cloud,” says Parameshwaran.  “The complexity of accessing and using these resources is hidden by Compute Manager.”

Whether or not HyperWorks On-Demand becomes a permanent member of Altair’s portfolio remains to be seen.  It is, after all, an experiment as the company, like so many other ISVs, explores the promises and pitfalls of cloud computing.

In the meantime, HyperWorks On-Demand can be used as a test bed for SMMs and larger manufacturing enterprises to try Compute Manager in Altair’s cloud experiment.  And Compute Manger is here to stay.