AI in Faroe Islands
- AI in Faroe Islands: Server Configuration & Deployment
This article details the server configuration used to support Artificial Intelligence initiatives within the Faroe Islands. It's designed for new system administrators and developers contributing to our AI infrastructure. This deployment focuses on balancing performance, cost-effectiveness, and resilience given the unique geographical and logistical challenges of the region.
Overview
The Faroe Islands, an autonomous territory within the Kingdom of Denmark, is increasingly leveraging AI for applications in fisheries management, weather forecasting, infrastructure monitoring, and healthcare. This requires a robust and scalable server infrastructure. Our current setup utilizes a hybrid cloud approach, combining on-premise hardware for latency-sensitive applications and cloud resources for burst capacity and redundancy. Data sovereignty is a critical consideration, driving the on-premise component. Network latency is also a major factor, influencing our choice of server locations and network providers. More information about our overall IT Infrastructure can be found on the main IT page.
Hardware Specifications
The core on-premise AI processing is handled by a cluster of servers located in a purpose-built data center in Tórshavn. Redundancy is built in at every level, from power supplies to network connections.
| Component | Specification | Quantity |
|---|---|---|
| CPU | Intel Xeon Gold 6338 (32 cores, 2.0 GHz) | 8 |
| RAM | 512 GB DDR4 ECC REG 3200MHz | 8 |
| GPU | NVIDIA A100 80GB | 4 |
| Storage (OS/Boot) | 1TB NVMe SSD | 8 |
| Storage (Data) | 16TB SAS HDD (RAID 6) | 12 |
| Network Interface | 100 Gbps Ethernet | 2 per server |
| Power Supply | 2000W Redundant | 2 per server |
These servers utilize Red Hat Enterprise Linux as the operating system, chosen for its stability and security features. The data storage utilizes a RAID 6 configuration for data redundancy and fault tolerance.
Software Stack
The software stack is built around the PyTorch deep learning framework, with supporting libraries for data processing and model deployment. We also utilize TensorFlow for specific projects requiring its capabilities.
| Software | Version | Purpose |
|---|---|---|
| Operating System | Red Hat Enterprise Linux 8.8 | Server OS |
| Python | 3.9 | Primary Programming Language |
| PyTorch | 2.0.1 | Deep Learning Framework |
| TensorFlow | 2.12.0 | Deep Learning Framework (secondary) |
| CUDA Toolkit | 12.2 | GPU Acceleration |
| cuDNN | 8.9.2 | Deep Learning Primitives |
| Docker | 20.10.17 | Containerization |
| Kubernetes | 1.27 | Container Orchestration |
Containers are managed using Docker and orchestrated with Kubernetes to ensure scalability and portability. We have a dedicated CI/CD pipeline for automated model deployment. Access to the servers is controlled via SSH and managed through a centralized authentication system.
Cloud Integration
For burst capacity and disaster recovery, we integrate with Amazon Web Services (AWS). Specifically, we utilize:
| Service | Instance Type | Purpose |
|---|---|---|
| EC2 | p4d.24xlarge | Backup AI Processing |
| S3 | Standard | Data Backup & Archiving |
| RDS | PostgreSQL | Model Metadata Storage |
| Lambda | N/A | Serverless Functions (Data Preprocessing) |
Data synchronization between the on-premise cluster and AWS S3 is handled by rsync over a dedicated high-bandwidth connection. The cloud resources are used for model training on large datasets and for providing failover capabilities in the event of an on-premise outage. Our disaster recovery plan is detailed in the Disaster Recovery Documentation.
Networking and Security
The server cluster is protected by a multi-layered security architecture. This includes firewalls, intrusion detection systems, and regular security audits. VPN access is required for remote administration. We adhere to strict data privacy regulations. Network monitoring is performed using Nagios to ensure system uptime and performance. All communication is encrypted using TLS/SSL.
Future Expansion
Planned future expansion includes upgrading the GPU infrastructure to NVIDIA H100 GPUs and integrating with a local Internet Exchange Point to reduce network latency. We are also exploring the use of federated learning techniques to improve model accuracy while preserving data privacy.
Server Monitoring is ongoing to ensure optimal performance and stability.
Intel-Based Server Configurations
| Configuration | Specifications | Benchmark |
|---|---|---|
| Core i7-6700K/7700 Server | 64 GB DDR4, NVMe SSD 2 x 512 GB | CPU Benchmark: 8046 |
| Core i7-8700 Server | 64 GB DDR4, NVMe SSD 2x1 TB | CPU Benchmark: 13124 |
| Core i9-9900K Server | 128 GB DDR4, NVMe SSD 2 x 1 TB | CPU Benchmark: 49969 |
| Core i9-13900 Server (64GB) | 64 GB RAM, 2x2 TB NVMe SSD | |
| Core i9-13900 Server (128GB) | 128 GB RAM, 2x2 TB NVMe SSD | |
| Core i5-13500 Server (64GB) | 64 GB RAM, 2x500 GB NVMe SSD | |
| Core i5-13500 Server (128GB) | 128 GB RAM, 2x500 GB NVMe SSD | |
| Core i5-13500 Workstation | 64 GB DDR5 RAM, 2 NVMe SSD, NVIDIA RTX 4000 |
AMD-Based Server Configurations
| Configuration | Specifications | Benchmark |
|---|---|---|
| Ryzen 5 3600 Server | 64 GB RAM, 2x480 GB NVMe | CPU Benchmark: 17849 |
| Ryzen 7 7700 Server | 64 GB DDR5 RAM, 2x1 TB NVMe | CPU Benchmark: 35224 |
| Ryzen 9 5950X Server | 128 GB RAM, 2x4 TB NVMe | CPU Benchmark: 46045 |
| Ryzen 9 7950X Server | 128 GB DDR5 ECC, 2x2 TB NVMe | CPU Benchmark: 63561 |
| EPYC 7502P Server (128GB/1TB) | 128 GB RAM, 1 TB NVMe | CPU Benchmark: 48021 |
| EPYC 7502P Server (128GB/2TB) | 128 GB RAM, 2 TB NVMe | CPU Benchmark: 48021 |
| EPYC 7502P Server (128GB/4TB) | 128 GB RAM, 2x2 TB NVMe | CPU Benchmark: 48021 |
| EPYC 7502P Server (256GB/1TB) | 256 GB RAM, 1 TB NVMe | CPU Benchmark: 48021 |
| EPYC 7502P Server (256GB/4TB) | 256 GB RAM, 2x2 TB NVMe | CPU Benchmark: 48021 |
| EPYC 9454P Server | 256 GB RAM, 2x2 TB NVMe |
Order Your Dedicated Server
Configure and order your ideal server configuration
Need Assistance?
- Telegram: @powervps Servers at a discounted price
⚠️ *Note: All benchmark scores are approximate and may vary based on configuration. Server availability subject to stock.* ⚠️