Numascale

Numascale is a Norwegian company that develops hardware and system software for scalable shared-memory computing systems targeted at high-performance and data-intensive workloads.

• High-capacity shared-memory system architectures for scale-up and scale-out computing
• Interconnect and node controller technology for cache-coherent Non-Uniform Memory Access (NUMA) environments
• System software and tools for configuring, managing, and monitoring large shared-memory clusters
• Solutions for technical computing, analytics, and simulation workloads in enterprise and research data centers
• Consulting and integration support for deploying and tuning shared-memory infrastructures

More About Numascale

Numascale focuses on scalable shared-memory architectures for organizations that run High performance computing (HPC) and data-intensive workloads and prefer a single system image over distributed-memory cluster designs. Its technology enables multiple industry-standard server nodes to be connected into a cache-coherent NUMA system in which all memory is addressable from any processor, with the interconnect providing coherency and access control. This architecture is relevant for applications and middleware stacks that expect large addressable memory spaces and benefit from direct load/store semantics instead of explicit message passing.

The company’s offerings center on hardware components and associated firmware and system software that implement a cache-coherent interconnect layer between server nodes. In a typical deployment, Numascale hardware is installed in or alongside x86-based servers and links them into a single logical machine with expanded memory capacity and core counts. The technology is aligned with symmetric multiprocessing (SMP) and NUMA concepts, with directory-based cache coherence protocols, memory-mapped communication, and support for standard operating systems that recognize the combined configuration as one shared-memory server.

For enterprise and institutional environments, Numascale positions its systems for workloads such as numerical simulation, modeling, in-memory analytics, and other technical computing use cases where application scaling can be constrained by inter-node message latency in traditional clusters. By extending the size of a single shared-memory node, organizations can run larger problem sizes per instance, simplify application porting from smaller Service Management Platform (SMP) systems, and use existing threading and shared-memory programming models such as Open Multi-Processing (OpenMP) and POSIX threads. This approach also aligns with software that benefits from very large in-memory datasets without re-architecting for distributed memory.

Numascale provides system software and management tools that support topology discovery, configuration, and monitoring of the shared-memory domain, integrating with standard Operating System (OS) facilities for NUMA awareness and resource management. These tools are used by system administrators and infrastructure teams to manage partitioning, performance tuning, and fault diagnostics. The company also offers integration and consulting services to assist with system design, installation, and workload characterization, helping customers align Central Processing Unit (CPU), memory, and interconnect capacities with application behavior.

Within an enterprise IT and HPC taxonomy, Numascale can be categorized under shared-memory computing infrastructure and interconnect technology (HPC infrastructure), with solution focus areas that include high-memory SMP/NUMA systems for technical computing, in-memory analytics platforms, and specialized hardware-assisted scaling for single-system-image environments. Its offerings are typically deployed in research institutions, engineering organizations, and data centers that maintain on-premises (on-prem) HPC or large-memory infrastructures and are looking to extend the capabilities of standard server platforms using shared-memory scaling approaches.