From November 17 to 22, 2024, Makat Tlebaliyev, a Computer Engineer at the Institute of Smart Systems and Artificial Intelligence (ISSAI), visited the International Conference for High Performance Computing, Networking, Storage, and Analysis at the Georgia World Congress Center in Atlanta, Georgia, USA. This event is the largest global conference dedicated to high-performance computing (HPC), networking, storage, and analysis, attracting a diverse audience from academia, industry, and government sectors. Throughout the conference, Makat Tlebaliyev actively engaged in sessions, presentations, and exhibitions while attending talks by top expert speakers.
On the first day of the conference, the opening ceremony was followed by tutorial sessions, including “AI and Scientific Research Computing with Kubernetes”. This session aimed to educate AI and computational science researchers on using Kubernetes as a resource management system, comparing it with traditional systems. Makat received an overview of Kubernetes architecture and job submission procedures, learned about storage options, conducted hands-on exercises with AI inference, training, and scientific research software using Kubernetes on CPU and GPU resources, and explored MPI examples. He also attended the sessions on “High-Performance and Smart Networking Technologies for HPC and AI” and “High-Performance Object Storage: I/O for the Exascale Era”, presented by Adrian Jackson from the University of Edinburgh’s EPCC, Dean Hildebrand from Google, Mohamad Chaarawi from Intel Corporation, and others. During this session, he learned about the design and usage of object stores as alternatives to traditional filesystems, using Ceph and DAOS as examples through hands-on exercises.
On November 18th, Makat attended a session on “High-Performance and Smart Networking Technologies for HPC and AI”, presented by speakers from Ohio State University. The session covered networking architectures, current market trends, and their suitability for designing HPC systems.
On November 19th, Makat attended the presentation “Lustre Community BoF: Lustre in HPC, AI, and the Cloud” by Peter Jones of Whamcloud and the European Open File System Association (EOFS). The discussion focused on the Lustre file system, a leading open-source solution for HPC. Makat met with Lustre developers, administrators, and solution providers, who shared recent system developments and challenges, such as Lustre’s role in AI and its use in cloud environments.
Next day, Makat was at the meeting titled “Nazarbayev University – DDN: Roadmaps and Plans”. Data Direct Network’s managers, engineers, and key customers discussed future improvements in storage solutions. James Coomer, Senior Vice President for Products, presented new AI storage solutions for NVIDIA AI clusters. The following day, Makat attended the presentation “InfiniBand’s Pivotal Role in Shaping the Future of Discovery” by Gilad Shainer from NVIDIA Corporation, Peter Salanki from CoreWeave, Sergio Iserte from the Barcelona Supercomputing Center (BSC), and Scot Schultz from NVIDIA Corporation. The presentation covered the convergence of AI and high-performance computing (HPC) and significant reductions in time-to-insight.
On November 21st, he joined an exhibit session called “The latest technologies, products, solutions, and services from cutting-edge innovators”, by keynote speakers Mr. Kevin Hayden from Argonne National Laboratory (ANL) and the University of Chicago. Makat also attended a customer presentation at the Signia by Hilton Atlanta titled “Supercharge Your Path from HPC to AI”, organized by ISSAI’s business partner Data Direct Network, and supported by NVIDIA. In the presentation, changes and future plans for all-flash storage solutions for Generative AI with NVIDIA Clusters were demonstrated.
On the last day of the conference, Makat participated in the presentation titled “Analyzing Parallel I/O” by Shane Snyder from Argonne National Laboratory (ANL), along with session leaders Jean Luca Bez from Lawrence Berkeley National Laboratory (LBNL) and Julian Kunkel from the University of Gottingen (GWDG). The presentation focused on best practices for identifying parallel I/O performance bottlenecks in applications.
The conference provided a valuable opportunity for Makat to get new expertise and knowledge. Makat learned best practices from engineers and academia in the USA’s AI and HPC sectors, connected with experts and colleagues worldwide, built new relationships, and joined a global community of professionals. ISSAI management is committed to supporting team members in their pursuit of continuous learning and professional development.