Intel, Google Cloud Optimize Performance for HPC Workloads

Estimated reading time: 2 minutes

Intel and Google are working together to drive high performance computing (HPC) forward on Google Cloud with the release of the Cloud HPC Toolkit. This new resource provides access to tools from the Intel® oneAPI Base and  HPC Toolkits to optimize performance through Intel® Select Solutions for Simulations & Modeling.  These new tools improve compile times, speed of results and offer multi-vendor acceleration in SYCL.

“Using Cloud HPC Toolkit with an Intel Select Solutions for Simulations & Modeling blueprint brings the added benefit of automatically spinning up a hardware-software configuration that has been rigorously tested and optimized for real-world performance, eliminating guesswork,” says Ilias Katsardis, HPC solution lead, Google.

Intel supports an open ecosystem to further industrywide innovation. Enterprises and research organizations are eager to adopt cloud computing to expand computing capacity and access the latest technologies for large, complex projects. Despite the desire to set up robust HPC environments in the cloud, many face challenges with overcoming unfamiliar concepts and tools. As a result, demanding workloads and applications are slow to deploy, encounter software incompatibilities and showcase subpar performance.

Intel Select Solutions for Simulations & Modeling also provides access to tools including the Intel® MPI Library and Intel® oneAPI Math Kernel Library via the Spack package manager, a popular software delivery and build platform. The culmination of work between Intel and Google helps make HPC more seamless and turnkey for users in this community.

The integrated Intel oneAPI tools support the open standards HPC developers need including C++, SYCL, Fortran, OpenMP, MPI and Python. They deliver productive cross-architecture performance across CPUs and GPUs using a single codebase.

The Cloud HPC Toolkit with Intel Select Solutions was developed using 2nd Gen Intel® Xeon® Scalable processors. It enables users to simplify configuration and deployment of HPC workloads and optimize for real-world performance. Intel’s ongoing collaboration with Google supports these projects, delivering more automated, user-friendly and compatible solutions that enable users to quickly meet increasing performance demands.

In August 2021, Intel announced a collaboration with Google that utilized Google’s HPC VM image and the Intel Select Solutions for Simulation & Modeling specification to automate the creation of Intel Select Solution-compliant HPC clusters on Google Cloud. Prior to that, Google worked with Intel to extend the capabilities of its CentOS-based VM image for HPC with the Intel HPC Platform specification, which made Intel’s software optimizations for Intel Xeon Scalable processors easily available to user applications running on Google Cloud.

Intel is extending open programming solutions into cloud environments to help drive innovation across the industry and adoption of new HPC CPU and GPU architectures as they become available. As demand for high-performance storage grows in HPC, so does the need for Distributed Asynchronous Object Storage (DAOS). As the foundation of the Intel exascale storage stack, DAOS is an open source software-defined scale-out object store that provides high-bandwidth, low-latency and high I/O operations per second storage containers to HPC applications. With this announcement, capabilities for DAOS are now hosted in the Cloud HPC Toolkit for a fully automated user experience.