Microsoft Azure Developers at SC22

Supercomputing (SC), the international conference for high-performance computing, networking, storage and analysis, is one of the biggest events of the year for those in high-performance computing. The 2022 event took place in Dallas, Texas, from the 13th to the 18th of November, and welcomed over 11,000 attendees to over 350 exhibit booths hosting researchers, scientists, application developers and more. Among the leading exhibiting companies, Microsoft had a strong presence with both technical sales and engineering Azure HPC specialists present to show off their latest product releases. 

What is High-Performance Computing?

High-performance computing (HPC) is the concept of using parallel data processing to improve computing performance and perform complex calculations. HPC does this by aggregating computing power, so that even advanced applications can run efficiently, reliably and quickly. It delivers much higher power and better performance than traditional computers, meeting the advanced needs of today’s businesses. Ultimately, groundbreaking discoveries often rely on technology, data and advanced computing. With cutting-edge technologies like artificial intelligence, machine learning and the internet of things (IoT) becoming more mainstream, there is more data than ever to be managed, processed and manipulated. HPC systems can perform quadrillions of calculations per second compared to the 3 billion calculations per second of a regular computer with a 3 GHz processor. 

What is Microsoft Azure’s HPC Offering? 

Azure high-performance computing (HPC) is a complete set of computing, networking and storage resources for HPC applications. The workload orchestration services come with purpose-built HPC infrastructure and solutions. What’s more, Azure incorporates next-generation machine learning tools to help drive smarter simulations and empower intelligent decision-making. Some of the key benefits of Azure’s HPC offering include:

• Optimised performance – users can leverage the entire spectrum of CPU, GPU, FPGA and fast interconnect capabilities while controlling costs at all times. 
• Workflow agility – users can build and manage their own HPC clusters to ensure end-to-to-end application lifecycle management in the cloud. 
• Production-class platform – users benefit from robust HPC reliability, data security and global regulatory compliance. 
• Built-in intelligence – users can incorporate intelligence into their solutions by building and training new AI models with automated machine learning, auto-scaling cloud compute and built-in DevOps. 

What Did Microsoft Do at SC22?

Microsoft held a leading role at SC22 announcing its newest product releases and latest work in the HPC space and highlighting partners who’ve leveraged Azure to unlock innovation. As well as having its own stand, it also partnered up with NVIDIA, the multinational technology company and inventor of the GPU, to deliver special presentations at a shared booth space. There was also an after-hour event hosted with AMD.

Within Microsoft’s booth, the company displayed hardware for some of its newest virtual machines for visitors to see up close, most notably the hardware powering its HBv4 instances. Microsoft was one of the first companies that discovered HPV could be done in the cloud, building the Infiniband-backed infrastructure with high-end compute nodes that would make HPC possible. HBv4 is Microsoft’s AMD EPYC 9004 Genoa instance hardware. Along with its two large processors, the incredible cooling centrepiece impressed visitors with heat pipes routing from both CPUs stretching across almost the entire chassis. Microsoft also showed off a large OAM chassis containing AMD Instinct MI250X OAM modules, with a huge array of heatsinks. Many of these technologies have gone from concept to reality in just a few short years. 

Key Sessions and Workshops

As well as the hardware on display, Microsoft also offered many sessions and workshops to give visitors a greater insight into the world of HPC. Some of the key sessions included:

• Kent Altena – Intel processors enabled an on-premises financial modelling ISV to make an Award winning HPC SADD solution running on Azure.
• Rob Futrick – have it your way: use the HPC job scheduler of choice in Azure.
• Rajani Janaki Ram – HPC accelerator.
• Hugo Affaticati – Azure scales 530B parameter GPT-3 Model with NVIDIA NeMo-Megatron.
• John Lee and Davide Vanzo – Scaling AI with Azure supercomputers
• Brian Lepore – Azure HPC file systems overview.

Meanwhile, two fully-attended workshops gave attendees a chance to deep-dive into the details of Azure’s HPC offering:

• Azure HPC and AI – the session dived into the details of Azure HPCs set of purpose-built computing, networking and storage resources and services for HPC applications. It explained how users can onboard quickly by building a highly-optimised production-ready Azure HPC environment and demonstrated a complete HPC cluster solution.
• Essential leadership skills for HPC – this half-day workshop delivered a quick tour through HPC delivery models, strategic considerations, metrics, value reporting and more. Especially relevant to academic, government and commercial sectors, the practical experience-based learning gave a unique insight into the business and leadership aspects of HPC. 

In addition, Microsoft continued to support the 2022 Student Cluster Competition. Microsoft collaborated with the SCC committee back in 2017 to introduce a cloud component to the competition, giving teams access to an Azure CycleCloud environment. This year’s competition saw teams supplementing their physical clusters with HPV offerings from Azure, such as HC, HBv3 and NDV4 VM types. 

Microsoft’s Commitment to HPC

Azure’s market-leading vision for HPC is based on Microsoft’s commitment to continual improvement and a wealth of recognised HPC expertise. The company uses proven HPC technology and design principles and enhances them with the best features of the cloud. As a result, Azure is able to deliver unprecedented performance, scale and value to its customers. Customers are able to rent supercomputer scale for their AI workloads and get the massive compute power needed to drive innovation relating to the environment, public health, energy, sustainability, economic growth and more. Ultimately, Azure’s HPC and AI cloud services deliver the agility, scale, security and leading-edge performance needed to access HPC and drive innovation forward.