Case study

Project Cerberus

Project Cerberus Overview

Intelligent Design, Accessible Power

Cerberus has been meticulously designed to slash cost overheads and make AI, ML, & HPC accessible to everyone. We believe the benefits of these technologies should not be exclusive. Our optimised approach starts with a powerful yet cost-effective hardware foundation:

  • Server Nodes: 3x Dell PowerEdge R750s (2U)
  • Processors: Dual Intel Platinum 8352Y per node
  • Memory: 256GB per node
  • Storage: 20TBs of NVMe SSD per node
  • Acceleration: Dual NVIDIA A40 GPUs per node

This highly capable 3-node cluster comes at a hardware cost of circa £90,000 (approx. £30k/node), a stark contrast to traditional outlays.

Optimised Performance & Flexibility

Optimised for Performance & Flexibility

CodeZero offers tailored software stacks to meet diverse needs and budgets:

1. Modest & Cost-Effective Stack:
  • Ubuntu Linux VMs, Docker containers, Prometheus, Grafana.
  • Enables AI, ML, & HPC workloads with GPU passthrough (one VM per GPU).
2. Enterprise-Grade Stack:
  • VMware vSphere (enabling vGPU for shared GPU resources)
  • VMware vSAN (Software-Defined Storage)
  • VMware Tanzu (Containers & Kubernetes)
  • VMware Horizon (VDI)
  • VMware Aria Operations (Monitoring)
  • NVIDIA AI Enterprise Suite

This comprehensive enterprise stack delivers robust functionality for mission-critical applications.

Innovative Deployment

Innovative Deployment

The Reverse Centralised Microservices Concept

The 3-server Cerberus cluster allows us to flex our innovative "Reverse Centralised Microservices" concept:

  • Node 1: Dedicated to GPU-based model training.
  • Node 2: Focused on inference (post model training).
  • Node 3: Provides VDIs for Data Scientists and Engineers.
Natively Designed for Immersion Cooling

The beauty of Cerberus lies in its native design for Immersion Cooling. This forms a strategic collaboration with industry leaders:

  • GRC (Green Revolution Cooling)
  • Castrol ON (part of bp Castrol)
  • Evoque-Cyxtera (one of the largest data centre colocation providers)

This approach ensures optimal thermal management, energy efficiency, and sustainability.

Project Cerberus in Action

The Impact: Unlocking Potential for All

Project Cerberus is more than a hardware and software configuration; it's a blueprint for democratising access to cutting-edge computational power.

By drastically reducing TCO and offering flexible, optimised solutions, CodeZero empowers businesses and institutions of all sizes to harness the transformative potential of AI, ML, and HPC.

Since its inception, Cerberus has gained international recognition for pushing the boundaries of what's possible with immersion cooling. The project has shown that complex computational tasks can benefit significantly from immersion cooling technology while maintaining exceptional energy efficiency.

With support from industry leaders like Dell, Intel, GRC, and Castrol ON, Project Cerberus continues to influence the direction of high-performance computing infrastructure design and implementation globally.

Ready to democratize access to AI, ML, and HPC?

Let's talk about how immersion cooling can make cutting-edge computing accessible and sustainable for your organization.

Contact us