Russian scientists will bring a massive new particle accelerator complex online next year, striving to gain a deeper understanding of the split-seconds following the Big Bang. The Russian Federation-led megaproject may also provide insights that can eventually help improve proton therapies for cancer treatment and process waste more effectively from nuclear power plants, among other potential applications.
And an innovative, composable HPC solution built on Intel compute and storage technologies will power the data acquisition and analysis for the accelerator project, which is based at the Joint Institute for Nuclear Research (JINR) in Dubna, Russia.
Solutions Focus
The solution was developed by RSC Group, working closely with Intel as a Titanium member of the Intel Partner Alliance Program. RSC is a Russia-based global developer and integrator of ultra-high-density, energy-efficient, innovative solutions for HPC, enterprise data centers, and cloud platforms. The company offers certified Intel Select Solutions for Simulation and Modeling and for Professional Visualization, as well as creating custom solutions such as the one at JINR.
“Our close working relationship with Intel technology experts helps us create leading-edge HPC solutions for customers in Russia and around the world,” says Alexander Moskovsky, CEO of RSC Group. “We can provide outstanding HPC solutions for the global scientific community and for manufacturing, finance, and other industries. This cooperation provides winning results for RSC Group and our customers.”
Hyperconverged Platform
RSC architected the JINR solution to handle the complex computational and data access requirements of diverse HPC workloads. These include high-energy and low-energy physics programs at JINR, as well as supporting the 26-nation Nuclotron-based Ion Collider Facility Accelerator (NICA) collaboration. The solution also had to operate within stringent cost, floor space, and thermal constraints.
The result is a hyperconverged platform that integrates Intel® Xeon® Scalable processors, Intel Optane™ persistent memory, Intel Optane SSDs, Intel NVMe SSDs, and Distributed Asynchronous Object Storage (DAOS), among other technologies.
“JINR needs very high-performance infrastructure for solving a wide range of theoretical physics tasks, as well as for modeling and processing experimental data,” says Andrey Moshkin, a software engineer at JINR. “The NICA experiments themselves will produce an expected data flow of up to 50GB/sec, and the total volume of incoming data per year can reach 1 Exabyte. To reduce overall storage requirements, we will do as much data processing as possible in real time, so we need the most efficient processors, at least 6GB of RAM per core, and the highest possible access speeds.”
RSC Group has worked with JINR for three years, creating JINR’s current supercomputer, named Govorun, in 2018, and upgrading it in 2019. Govorun is a cost-effective, high-density supercomputer that uses RSC’s patented, hot-water-cooled technology to achieve an impressive average power usage efficiency (PUE) of less than 1.04.
Reconfigurable on the Fly
Govorun combines 2nd Gen Intel Xeon Scalable processors with a distributed, hierarchical storage-on-demand architecture. The top layer of the storage hierarchy uses Intel Optane persistent memory (Intel PMem) and DAOS to store metadata and handle real-time processing of sensor data from the NICA experiments, along with other tasks requiring the highest processing speeds. Other workloads can use Govorun’s NVMe-over-fiber drive pool of Intel Data Center SSDs.
The solution also includes RSC BasIS, a unique software interface and orchestration engine that provides storage-on-demand flexibility with highly granular controls. Users can create multi-layered, composable storage systems on the fly, dynamically selecting, combining, and configuring drives to match the user’s workload types—all within seconds. Govorun users can switch among Lustre file systems, parallel Network File Systems (NFS), and distributed DAOS storage systems.
A pilot implementation at JINR using RSC’s Tornado server platform, which includes 2nd Gen Intel Xeon Scalable processors and Intel Optane persistent memory 100 series, held 27th place on the July 2021 on the IO500 global rating, making it the highest-ranked Russian system on the list.
JINR and RSC Group are planning to upgrade Govorun to 3rd Gen Intel Xeon Scalable processors and expand the amount of Intel PMem and Intel SSDs to create a 1.6PB DAOS. “By using 3rd Gen Intel Xeon Scalable processors with Intel PMem to extend our RAM capacity, we expect to increase the speed of NICA calculations that require a large amount of RAM up to two times,” says Dmitry Podgainy, leading researcher of JINR. “We are counting on RSC and Intel to keep us moving forward as we advance into an exciting new era of scientific exploration.”
Learn more about Intel® Select Solutions.
Intel technologies may require enabled hardware, software, or service activation.
Intel does not control or audit third-party data. You should consult other sources to evaluate accuracy.
© Intel Corporation. Intel, the Intel logo, and other Intel marks are trademarks of Intel Corporation or its subsidiaries. Other names and brands may be claimed as the properties of others.