A summary of high-performance computing (HPC) that focuses on simplifying the concept and explaining with the help of examples.
In the recently concluded #Rethink360Cloud Virtual Summit, I presented a session on High-Performance Computing (HPC) and why it should be used to solve complex problems quickly. The focus was to simplify the concept and explain with the help of examples.
The session is based on IBM Cloud and HPC webinars that have been conducted previously by our SMEs. This blog post has a summary of my session notes and tips to help you experience the power of HPC through a simulation model.
What is high-performance computing (HPC)?
To understand high-performance computing, let’s start by considering a simple scenario. Which of these samples do you think will complete the task first: One person performing 10,000 calculations or 10,000 people collectively performing 10,000 calculations?
The parallel that I draw is this—the more thinking and doing power you have, the faster the work shall be done.
This can be the basis of understanding what HPC is. Known by many names over its evolution—machine learning, grid computing, deep learning, distributed learning, distributed computing—HPC is basically when you apply a large number of computer assets to solve problems that your standard computers are unable or incapable of solving.
This could be because they are not equipped with the right resources, take too long to run or, are unable to handle the load (think faster, larger—in bursts or, over long periods).
If you look at it, you are basically handling “large scale, parallel processing” workloads that can run large analytical computations—e.g., millions of scenarios, using up to terabytes data (weather forecasts, risk management).
Why do you need HPC?
High performance is all about being able to leverage immense computing power to scale your business and its innovations to the next level. On top of that, if you can get the flexibility that is inherent to cloud—on-premises, hybrid, or public cloud—then you are in a true growth mode. Another key reason you should look at HPC for your critical computing needs is the ability to dynamically handle workloads, keeping a check on your costs as well as performance requirements. If security is a key concern for your business, look at the additional layers that HPC brings—bring your own key, encrypted OS, or leverage a hardware-level security module.
Where can HPC be applied?
High-performance computing (HPC) in the cloud presents opportunities across industries. You can basically build your personalized end-to-end HPC solution tailored to your organization’s needs. To understand this a bit more, let’s look at some examples where HPC can be beneficial:
Intense or highly complicated calculations and simulations
HPC is critical for any applications where you have mathematically intense or highly complicated calculations. Consider engineering or scientific research, where the teams can run complex simulations against large datasets. Other excellent examples would be in fields such as aerodynamics, physics, or pharmaceuticals. The ability to handle complex predictive analytics also means the reduced cost of operations for the business.
Another great example would be building simulation models (say, for banking and financial services) by leveraging AI and machine learning capabilities. There is a high degree of accuracy that HPC can bring in when performing large-numbered, complex financial transactions
Faster and more accurate design iterations for new products
The faster an engineering team can iterate on a design, the faster the product, service, or solution can be brought to market. Plus, this provides a significant competitive advantage. A great example would be modeling capabilities and analysis of data structures in the aerospace or automotive industries. By being able to create and evaluate virtual prototypes in real-time, you can identify possible problem areas and then iterate on the design, all before you even get to production. Not only are you reducing your costs and the resources involved, you are bringing better products to market and increasing business credibility.
Anytime you require large amounts of compute resources to complete a task in a reasonable period of time, HPC is a good option because it brings in the necessary speed and scalability. For the film, media, or gaming industries, graphic design is a critical part of the entire production process. HPC can help create smoother and richer multimedia applications (including 3D-intensive programs), render complex architectural models faster, and streamline animations for different delivery mediums—television, cinema, or mobile devices.
Another great application is in life sciences. With studies around genetics advancing at a tremendous rate, HPC helps in critical operations like mapping genomes, processing them, and sequencing them based on models created—all in days instead of the weeks or months that such complex models might need. Imagine faster drug design and production for all the new diseases that seem to pop up in the news, almost every other day.
Resource balancing and cost optimization
This is highly applicable in today’s world, with diverse and remotely located businesses. When delivered in a cloud environment, product development teams can be located anywhere in the world because compute resources need not be spread throughout every company location. You could do inventory analysis, logistics and supply chain optimization, sentiment analysis, and create marketing offers for your retail business, all while located remotely.
You can also make intelligent and informed decisions with remotely located resources, like a cross-geo intelligence agency. Conducting fraud analysis, climate modeling, weather forecasting, energy, nuclear stewardship, exploration—especially for government and defence requirements—becomes easier if you can get the intense power of HPC with the benefits of being on the cloud.
What deployment options are available?
IBM has a number of deployment solutions for high-performance computing needs.
These can process large volumes of data more economically and quickly with an easily configurable and scalable solution on the IBM Cloud. You can meet your workload demands, choose high-speed data transfer and storage that expands to exabytes, and leverage industry-leading workload management software.
IBM Spectrum Computing is new and intelligent resource and workload management software that is designed to make it easier for organizations to extract full value from data to accelerate performance-intensive analytics or machine learning. This technology can be used across industries to help sequence genomes for improved cancer treatment, allow engineers to design a championship-winning Formula One race car, or let bankers personalize financial services to attract new customers:
Spectrum Computing LSF: IBM Spectrum LSF® enables easy deployment, maximum uptime, and performance at scale—critical to succeed with HPC—increasing user productivity and hardware utilization while decreasing system management costs. You can accelerate research, design, and other high-performance simulations by up to 30% with comprehensive workload management.
Spectrum Computing – Conductor: Share clusters and accelerate modern, distributed workloads up to 58%, including Spark analytics, AI frameworks, and containers. Confidently deploy Spark and other services in a multitenant environment, both on-premises and in the cloud.
Spectrum Computing – Symphony: Supercharge analytics with a low-latency, high throughput SDI that delivers 700x greater scalability and 100x faster results. Benefit from enterprise-class management for running compute- and data-intensive distributed applications on a scalable, shared grid.
HPCaaS from Rescale on IBM Cloud is a cloud simulation platform that helps engineers and scientists build, compute, analyze, and scale simulations with high-performance computing. HPCaaS from Rescale enables you to rapidly deploy your HPC jobs on the IBM Cloud with your chosen execution environment. The service allows you to pick your desired software package and compute and storage configuration for your HPC job.
Once submitted, HPCaaSfrom Rescale provides a live tailing feature and enables easy access to the execution environment for debugging. Once your job has completed, you can download the results or continue to do further analysis on the IBM Cloud.
I hope our product demonstrations and product models will get you off to a good start with HPC. You can also hear our technical experts walk you through common use cases, highlighting product features key capabilities.