17:50, 13 March 2026

MIPT Launches Educational Computing Cluster at New Data Center

The new computing cluster at Moskovskiy fiziko-tekhnicheskiy institut (Moscow Institute of Physics and Technology) will support both research projects and hands-on training for students.

Moskovskiy fiziko-tekhnicheskiy institut (Moscow Institute of Physics and Technology) has launched an educational computing cluster based at the university’s new data center. The high-tech facility was commissioned in 2025 and will be used not only for scientific research but also for student training, the university’s press service told IT Russia.

GPU Accelerators and High-Performance Servers

The complex includes servers equipped with NVIDIA H200 graphics accelerators, high-performance CPU servers, and other computing hardware. The infrastructure covers part of the university’s demand for computing capacity used in academic programs. The cluster’s capabilities are expected to expand in the future.

The cluster will be used in courses run by the institute’s departments and fiztekh schools that train specialists in programming, high-performance computing, artificial intelligence, and mathematical modeling. The main goal is to give students the opportunity to work with infrastructure that closely matches modern industry standards.

According to MIPT Rector Dmitry Livanov, the cluster built on the university’s new data center is a logical continuation of the institute’s strategy to integrate modern IT infrastructure into education. Its creation marks an important step in developing the university’s digital ecosystem. Students will now be able to access high-end IT infrastructure without waiting for resources to become available.

“I am confident that direct access to such resources will help students engage with major scientific research more quickly and increase the number of high-quality academic papers they publish even before graduating,” Dmitry Livanov said.

SLURM at the Core

The first users of the cluster will be faculty members and students working on machine learning and computational physics projects. Access to the system will later be expanded.

The cluster operates under the SLURM workload manager, which automatically allocates resources while prioritizing educational tasks. Transparent usage limits have already been set for students and faculty—for example, up to 32 CPU cores and two GPUs per job. Continuous computing time is limited to 24 hours, which is sufficient for most academic workloads.

“When designing the cluster, it was important for us to maintain a balance between making resources accessible to students and ensuring the stability of the institute’s entire computing infrastructure, while also enabling flexible scaling without disrupting that balance. That is why access is organized through a transparent system of limits and requests,” said Ruslan Pirogov, head of the university’s IT department.

Data is stored across two tiers: a fast NVMe storage layer (1 TB) for active tasks and a slower storage layer (10 TB) for long-term data.

Earlier reports said Natsionalnyy issledovatelskiy yadernyy universitet MEPhI (National Research Nuclear University MEPhI) had developed a unique mass spectrometer as a domestic alternative to imported systems.

Education