러시아가 48코어 바이칼(Baikal) 칩으로 구동되는 자체 슈퍼컴퓨터 스토리지 시스템을 구축했다고 Frontier India가 보도했다.

Russia has completed deploying a domestically developed data storage system for the supercomputer infrastructure of Moscow State University (MSU), another step in integrating Russian-designed processors, server hardware, and software into high-performance computing (HPC). The SDS TROK software-defined storage platform, developed by Astra Group, runs on servers manufactured by Spintronika and powered by 48-core Baikal-S processors.

The installation is part of MSU's Istok hardware-software complex, which has been operating since 2026. The project was officially announced on November 9, 2026, after Astra Group and Spintronika completed development and integration work, with financial support from Russia's Ministry of Science and Higher Education. The system provides storage for scientific computing workloads and lets researchers access stored data via the Network File System (NFS) protocol.

The project is significant because the university specifically required compatibility with Russian processors and domestically developed components. The companies prioritized making the software compatible with Baikal-S-based server hardware rather than adapting an existing foreign storage platform.

**SDS TROK: A Storage Platform Built Around Russian Hardware**

Astra Group's SDS TROK is a software-defined storage platform. Unlike traditional storage systems, which rely on dedicated proprietary hardware, software-defined storage uses software to manage storage resources distributed across multiple servers. Administrators can consolidate hardware resources into a centrally managed storage environment and increase capacity by adding nodes.

For the MSU project, SDS TROK was implemented on Spintronika-designed and manufactured servers using Baikal-S processors. Users reach shared data through NFS, a standard network file-sharing protocol widely used in Linux environments. Researchers can use the storage without adopting a different data-access procedure.

The Ministry of Science and Higher Education supported the initiative. MSU ran a competitive program to secure funding for modernizing the material and technical infrastructure of organizations that conduct scientific research and development.

Integration was a critical part of the project. The team set up a dedicated software build and assembly pipeline and then ran two phases of testing. The first validated software compatibility using virtual machines. The second tested physical servers equipped with Baikal-S processors. The developers reported stable operation and adherence to the required specifications in both stages.

The main SDS TROK codebase has also been updated with the modifications developed for Baikal-S compatibility, so the enhancements are not limited to the MSU installation. Customers deploying the platform on Russian processor architectures can also use them.

**Baikal-S: The 48-Core Processor Behind the System**

The storage servers are built on the Baikal-S, also known as the BE-S1000. Baikal Electronics designed it for storage infrastructure, high-performance computing systems, and server platforms.

The processor implements the Armv8-A architecture with 48 ARM Cortex-A75 cores running at 2.5 GHz. It supports up to 768 GB of memory per processor and six DDR4-3200 memory channels with ECC. It offers up to 80 PCIe 4.0 lanes, which can connect high-speed storage controllers and other peripherals, and 32 MB of L4 cache. Its specified thermal design power is approximately 120 watts.

These characteristics make the Baikal-S especially suitable for infrastructure projects that require high memory bandwidth, connectivity, and parallel processing.

Processor compatibility is only one part of the engineering challenge. A storage platform must also support the operating system, storage-management software, network interfaces, and data replication mechanisms. The MSU deployment therefore shows the processor working as part of a complete infrastructure rather than as a standalone server.

**DRBD Replication Protects Scientific Data**

Data protection is one of the most important features of the SDS TROK installation. Scientific computing systems often run calculations for long periods, and losing stored results to a server failure can waste significant computing time and force researchers to repeat work.

SDS TROK addresses this by using Distributed Replicated Block Device (DRBD) technology for synchronous replication. A write is acknowledged only after the data has been stored on the required number of nodes. If a server fails, the replicated data remains accessible on the other participating nodes.

The system also includes automatic recovery. When a failed server returns to operation, the storage cluster starts delta synchronization, copying only the blocks that changed during the outage instead of the entire dataset. This reduces network traffic and lets the node rejoin the fault-tolerant configuration more quickly.

Responsibilities are divided among node types. Controllers handle the configuration and overall coordination of the storage environment, while worker nodes manage the DRBD resources. This design allows the system to scale while keeping administration centralized.

According to the developers, administrators do not have to manually manage replication or start recovery after each failure. Interruptions should be minimized for researchers accessing NFS volumes, and previously stored calculation results should be protected.

**Baikal-S Expands Its Role in Russian Infrastructure**

The MSU installation is part of a broader effort to establish Baikal-S as a platform for Russian server and infrastructure equipment, and several domestic hardware and software developers have expressed interest in its 48-core architecture.

Aquarius's S2-1 storage infrastructure, built on its AQ2BS1000 server with Baikal-S, is designed for protected IT infrastructure, including government organizations and critical information infrastructure. Graviton's S2124B server, equipped with two 48-core Baikal-S processors, is aimed at high-performance computing and artificial intelligence applications.

Astra Group has also used Baikal-S in other parts of its product line. Astra Cloud was commercially launched on the same processor architecture, and the Tantor XData 2B database machine was adapted for the company's internal 1C environment.

**Istok Moves Toward Full-Scale Scientific Computing**

The Istok complex has been operational at Moscow State University since 2026. Research teams from MSU and more than ten organizations affiliated with Russian Academy of Sciences institutes use it.

The system is built for computationally intensive research, including environmental studies, biomedical research, cognitive science, and climate modeling. These workloads generate large volumes of data that must be stored, accessed, and protected throughout the computing process.

The university's preliminary performance evaluations indicate that climate-modeling tasks are being completed significantly faster than on previously used computing resources of comparable size. Beginning in 2027, the university plans to move to full-scale scientific calculations and research projects.

As workloads grow, the storage system will take on a more critical role. Large-scale simulations require more than processor performance. They also need the capacity to preserve intermediate and final results, along with reliable storage and adequate network bandwidth. A failure in any of these areas can affect the efficiency of the entire computing operation.