ANSYS has sponsored this article.
In the world of engineering and design, high-performance computing (HPC) has been used to accomplish complex tasks with ease. This has led to its popularity, but many still struggle with figuring out the proper configurations, how, and when it should be deployed. This struggle is partly due to confusing HPC with traditional supercomputers and the actual concept behind it. This post will focus on describing the concept known as high-performance computing. At the end of the article, you will understand:
- The definition of HPC
- Why engineers and manufacturers need it
- How engineers can use it
- The myths hindering the adoption of HPC
- The HPC solutions currently on the market
What Is HPC?
Like most relatively new technical concepts, it takes time for adherents to come to a consensus where definitions are involved. This happens to be the case with high performance computing, but a general understanding of HPC exists. In short, HPC systems can be defined as the aggregation of computing resources to provide solutions to the complex problems a personal computer cannot solve.
This generic definition is why many still make a claim to define all HPC systems as supercomputers, but this is fundamentally wrong. The fact that personal computers can be integrated into a cluster to provide high-performance computing negates that argument. While supercomputers can deliver targeted HPC services, making them a niche example of an HPC system, not all HPC systems consist of supercomputers. A perfect example of this phenomenon is Cray supercomputers.
The early Cray computers of two to three decades ago where solely standalone supercomputers. These computers were equipped with customized features needed to handle complex simulation and calculations making them niche HPC solutions. Today, Cray computers have embraced HPC and now offer a cluster of supercomputers that embrace interconnected technologies in delivering high-performance computing.
Another important distinction to consider when evaluating supercomputers as HPCs, is cost and customization. Most supercomputers are built to run customized software applications, and the cost of obtaining and running one is in the millions of dollars. This clearly puts every supercomputer out of the reach of 90% of the world’s businesses. HPCs on the other hand, are interconnected computer systems that use legacy software and are affordable to use. This means the average HPC system consists of thousands of high-end personal computers dedicated to solving complex problems.
How High-Performance Computer Systems Function
Every computer in an HPC system is known as a node. Each node is generally equipped with multiple processors called compute cores that handle the computation aspect of problem-solving. The processors, graphical processing units, and memory of each node are then interconnected by a network for high-performance computing.
If you are familiar with 3D printing, you can compare HPC ecosystems to a 3D printing farm or cluster. Here, multiple devices work together to deliver a much larger output or result. When multiple individuals intend to use the HPC system, a scheduler comes into play. The scheduler allocates resources to multiple users and intuitively scales up allocations according to a user’s requirements. In terms of storage, HPC projects require extensive storage space. This is one of the reasons cloud-based HPCs were developed.
Why is HPC Needed?
In the early days of Bitcoin mining, HPC systems were used to do the job and this played a role in popularizing the concept within the SME crowd. Today, HPCs are mostly used in brick and mortar establishments for some serious problem-solving activities. In the engineering field, HPCs are used in handling complex simulations. These simulations are complex modeling phenomena that are small in scale and transient, such as acoustics or combustion phenomenon in engines.
In multi-physics simulations, the simultaneous effect of multiple (natural or artificial) phenomena on a product or structural design is studied. An example is analyzing how the effects of thermal stress affects a moving structure, as well as, the turbulence caused by the moving structure at particular moments. Here, paired partial differentiation and other complex mathematical modelling must be calculated to solve the problem. HPCs can speed up the process and drastically eliminate errors that could affect the structures ability to function.
In industrial design and manufacturing, HPCs have been used to solve structural and thermal design processes to optimize production lifecycles. An example, was Rolls Royce attempting to eliminate the use of thermal sensors in its prototype engines for future automobiles. To accomplish this task, Rolls-Royce intended to employ an iterative process to determine the heat flux and the effects of removing thermal sensors to the engine’s structure. The power systems giant turned to ANSYS Fluent and HPC Pack. Fluent’s CFD (computational fluid dynamics) solver was used to determine the heat flux and CPU 24/7 HPC handled the computational workload.
In the end, Rolls-Royce drastically reduced the time required for a coupled CFD-structural simulation by 80 percent. Other benefits included avoiding costly design revisions during the production of the prototype and successfully developing a pathway for the elimination of thermal sensors in its engines.
An example of generative design within an HPC system is Airbus’ attempt to revolutionize commercial airplanes. The idea was to develop a conceptual aircraft with reduced weight, and an increased compartment structure, while retaining the strength of its commercial planes. To accomplish this manufacturing objective, millions of variables and structural interactions were modeled simultaneously to get an optimal product. The magnitude of this computational task led Airbus to high performance computing. According to Gerd Buttner, HPC is helping Airbus re-imagine the future of flight.
HPCs can also be used to manage a complex manufacturing process that requires multiple industrial licenses throughout production. To accomplish this, a conducive systematic approach that takes care of licensing must be used. The ANSYS Parametric Pack is one such solution. The parametric pack enables the simultaneous analysis of diverse design points using a set of licenses at an affordable cost. This speeds up the production process and reduces the time spent analyzing and acquiring multiple analyses.
Challenging Popular Misconceptions Around HPC
HPCs have predominantly been used by large scale businesses in years past and this has led to various misconceptions or myths about its use or applicability. One of the misconceptions currently making the rounds is that these systems can only be used for super complex simulations, such as CFD. This may be due to the scale of CFD simulations and the complexity of the underlying equations. However, HPCs are also be used for structural and electromagnetic simulations.
Another myth revolving around its use for SMEs is cost. Most project managers and CTOs still believe HPCs are too high priced with a low ROI but the facts state otherwise. With the advent of cloud-based HPCs, high cost may not be an issue. Vendors such as ANSYS and Microsoft now offer scalable pricing models at a lower cost for complex engineering tasks. This means businesses can choose to pay for only what they use and scale up at need.
The last myth is centered around ease of use. Many believe a dedicated IT supervisor or manager is needed to handle HPC-related activities. Today, there are currently managed HPC systems you can choose to take advantage off.
Understanding the Options
The HPC space is rapidly expanding and becoming more user-friendly due to the diverse vendors offering integration services. These vendors provide built-in solutions which allow the use of HPCs within existing environments. The process of integration can either be through parallel file systems or a cluster management solution. Cloud computing also provides a more affordable avenue for businesses interested in exploring the benefits of HPC services. The advantages of HPC cloud-based services include a scalable, on-demand and inexpensive option for handling demanding simulations. This is why approximately 64 percent of HPC platforms now integrate cloud computing and the market is expected to grow annually by 12 percent within the next 5 years.
Today, ANSYS works with many HPC vendors that offer streamlined HPC services to SMEs and large corporations.
These vendors also offer post-sales services to ensure integration and deployment tasks are properly executed.
Looking Forward
The future of engineering and manufacturing may lead to more use of HPC. In the event Moore’s Law continues to hold true, this future may be defined by small businesses and entrepreneurs looking to tap into the potentials a cluster of smart devices holds. Today, as more vendors continue to step up their offerings of HPC as a Service either on-premise or in the cloud, its use will become accessible and affordable to all.
To learn more about HPC, visit ANSYS.