ZEKLARData
EN / IT
Hardware

MONOLITH

The industrial computing center.

MONOLITH is Zeklar's industrial computing center: dual-to-quad Xeon CPUs, up to 256 GB of RAM and GPU clusters, sized for heavy simulation and AI workloads. Available in Basic, Pro and Enterprise tiers, with 24/7 support and SLAs for critical infrastructure. It is the machine that runs the models and analyses an edge node cannot carry.

256 GB MAXIMUM RAM24/7 SUPPORT AND SLA4x Xeon CPU SOCKETSADVISORY MODE · READ-ONLY256 GB MAXIMUM RAM24/7 SUPPORT AND SLA4x Xeon CPU SOCKETSADVISORY MODE · READ-ONLY
01In figuresFIELD DATA
256 GB
Maximum RAM
24/7
Support and SLA
4x Xeon
CPU sockets
02Composition
What's insideRack compute center · Basic / Pro / Enterprise
01CPU02RAM03GPU04NVMe05NET06PSU

Hover a component

CPU01 / 06

CPU

2–4× Intel Xeon Silver

Up to quad-socket for heavy simulation and inference workloads, on-prem.

03The problem

Simulations that do not fit on an edge node

Thermal-hydraulic analyses, digital twins and AI model training demand dozens of cores and memory no field node holds. Renting public cloud for sensitive nuclear data is often not permissible, and engineering workstations do not scale. You need dense compute, on-premise and under control.

04What it does04
01

Multi-socket Xeon CPUs

Dual-to-quad Xeon processors for dense cores and high throughput on parallel workloads and numerical simulation.

02

Up to 256 GB RAM

Registered ECC memory for large datasets and meshes, with runtime error protection.

03

GPU clusters

GPU accelerators for inference, training and scientific computing, orchestrated to share CORTEX workloads.

04

24/7 support

Round-the-clock support with per-tier SLAs, hot-spare parts and proactive hardware monitoring.

05How it works04 STEPS
01

Size

Pick the Basic, Pro or Enterprise tier based on the cores, RAM and GPUs your workloads require.

02

Install

The system ships rack-ready, is integrated into the data hall and connected to the on-prem Zeklar network.

03

Orchestrate

FORGE and CORTEX jobs are scheduled onto the compute pool with priorities and quotas.

04

Scale

Add nodes or GPUs over time; 24/7 support keeps continuity through expansion.

06Architecture

MONOLITH is the on-prem compute center that runs the heavy Zeklar workloads: FORGE simulations, CORTEX training and inference. The architecture combines multi-socket Xeon CPU nodes, a GPU pool, high-performance storage and a scheduler that dispatches jobs with quotas and priorities, all inside the customer perimeter.

01Compute nodesDual-to-quad Xeon servers with registered ECC memory up to 256 GB per node, sized for meshes and datasets no edge node can hold.
02GPU acceleration poolGPU accelerators with high-bandwidth interconnect for CORTEX training and inference and scientific computing, shared across projects via the scheduler.
03Storage and fabricParallel NVMe storage for simulation scratch and capacity tiers for results, over a low-latency fabric (25/100 GbE or InfiniBand).
04Job orchestrationHPC scheduler with batch queues, quotas and per-tier priorities: FORGE and CORTEX jobs are dispatched to the pool under tracked policies.
05Management and supportProactive hardware monitoring, hot-spare parts and 24/7 support with per-tier SLAs, integrated with AEGIS for resource licensing and identity.
07Specifications
CPU2-4x Xeon Scalable
RAMfino a 256 GB ECC reg.
GPUcluster fino a 8 acceleratori
StorageNVMe scratch + capacità HDD/SSD
Network25/100 GbE o InfiniBand
PowerPSU ridondata 200-240 VAC
Form factorRack 19" 2U-4U
TiersBasic / Pro / Enterprise
08Deployment

Rack in data hall

Rack-ready installation in the customer's data hall, integrated with cooling, redundant power and the on-prem Zeklar network.

Scalable on-prem cluster

Multiple nodes orchestrated as an HPC cluster, expandable over time by adding nodes or GPUs without stopping running workloads.

Air-gapped enclave

Configuration isolated from any external network for sensitive nuclear data, with updates delivered only through signed, verified media.

09Security & compliance

Verified boot and firmware

Secure Boot and TPM 2.0 on every node, with BMC/BIOS firmware attestation before a node joins the cluster.

Data encrypted at rest

Scratch and archive volumes encrypted, with keys managed on-prem: simulation data never leaves the customer perimeter.

RBAC and job audit

Queue and data access by role via AEGIS; every job, dataset and result is tracked with author and history for audit.

Certifications

ISO 27001-aligned operation; suitable for air-gapped enclaves and sovereign accreditation schemes as the base for CANON OS.

10Integrations
FORGE for simulation and designCORTEX for AI compliance workloadsVAULT as the plant data archiveAEGIS for compute resource licensingStandard HPC schedulers and batch queues
11FAQ

Why on-prem instead of public cloud?

Because public cloud is often not permissible for sensitive nuclear data. MONOLITH keeps compute and data inside the customer perimeter, under their control.

How do I choose between Basic, Pro and Enterprise?

By the cores, RAM and GPUs your workloads require: Basic for single projects, Pro for teams, Enterprise for plant fleets with tighter SLAs.

Can I grow capacity over time?

Yes. You add nodes or GPUs to the cluster; 24/7 support keeps continuity through expansion without stopping active jobs.

How are FORGE and CORTEX workloads shared?

An HPC scheduler dispatches jobs to the CPU/GPU pool with per-tier quotas and priorities, so simulation and AI coexist without contending for resources.

What SLA does 24/7 support cover?

Proactive monitoring, hot-spare parts and per-tier response times, designed for critical infrastructure that cannot stop.

12The markGRID 64 · STROKE 5

The mark encodes the mechanism, not the sector: platform and infrastructure: the enabling layer..

44302216
13Other products · ZEKLAR Data

ZEKLAR software technology 4.0/5.0

Let's talk
ZEKLAR · SOFTWARE THAT MAKES THE DIFFERENCE.
ZEKLAR · VAT L81419031T · info@zeklar.com© 2026