System Requirements

​

Production Requirements

​

Operating System Support

​

Supported Distributions

​

Hardware Requirements

​

CPU

• 16+ cores (32+ threads)
• Intel Xeon or AMD EPYC
• AVX2 instruction set support
• 3.0 GHz+ base clock

• 32+ cores
• Latest generation processors
• AVX-512 support

​

Memory (RAM)

• 32 GB - A reasonable starting point for moderate concurrency
• 64 GB+ - For high-volume deployments

​

Storage

• 500 GB NVMe SSD - Recommended starting point
• 1+ TB NVMe SSD - For high-volume or long-term log retention
• RAID 10 configuration recommended for production

​

Storage Performance

​

GPU (Optional)

• NVIDIA T4 (16 GB)
• NVIDIA A10 (24 GB)
• NVIDIA A100 (40/80 GB)
• NVIDIA L4 (24 GB)
• NVIDIA L40 (48 GB)

• CUDA 11.8 or higher
• NVIDIA driver 520.61.05 or higher
• Compute capability 7.0+

• A GPU can substantially reduce analysis time for model-driven detection and OCR-heavy documents. The size of the improvement depends on the model, document mix, and CPU baseline, so benchmark before sizing around it.
• Recommended for high-volume deployments

​

Network Requirements

​

Bandwidth

• 10 Gbps - Recommended for production
• 10+ Gbps with redundancy - For high-volume enterprise deployments

​

Ports

​

Firewall Rules

# HTTP API
sudo ufw allow 8080/tcp

# HTTPS API
sudo ufw allow 8443/tcp

# Metrics (from monitoring network only)
sudo ufw allow from 10.0.0.0/8 to any port 9090

​

Software Dependencies

​

Container Runtime

# Ubuntu/Debian
curl -fsSL https://get.docker.com -o get-docker.sh
sudo sh get-docker.sh

# Verify
docker --version

# Required version: 2.0+
sudo apt-get install docker-compose-plugin
docker compose version

​

Database

• 10 GB minimum
• 50 GB recommended
• Scales with audit log retention

​

Messaging

​

High Availability Requirements

​

Compute

• Minimum: 3 nodes
• Recommended: 5+ nodes
• Load balancer: NGINX, HAProxy, or cloud LB

​

Database

• PostgreSQL: Primary + 2 replicas
• Replication: Streaming replication
• Automatic failover: Patroni or similar

​

Messaging

• NATS: Clustered (3+ nodes) for redundancy

​

Storage

• Shared storage: NFS, S3, or similar
• Replication: 3x minimum
• Backup: Daily with off-site replication

​

Cloud Provider Recommendations

​

AWS

• Minimum: t3.xlarge (4 vCPU, 16 GB)
• Recommended: c6i.4xlarge (16 vCPU, 32 GB)
• With GPU: g5.2xlarge (1x A10G GPU)

​

Azure

• Minimum: Standard_D4s_v5 (4 vCPU, 16 GB)
• Recommended: Standard_D16s_v5 (16 vCPU, 64 GB)
• With GPU: Standard_NC8as_T4_v3 (1x T4 GPU)

​

Google Cloud

• Minimum: n2-standard-8 (8 vCPU, 32 GB)
• Recommended: n2-standard-16 (16 vCPU, 64 GB)
• With GPU: n1-standard-8 + 1x T4

​

Capacity Planning

​

Concurrent Jobs

​

Storage Sizing

Total Storage = (Avg Document Size × Daily Documents × Retention Days) + Base Storage

Example:
(5 MB × 1000 documents × 30 days) + 50 GB = 200 GB

​

Benchmarking

# CPU benchmark
sysbench cpu --threads=16 --time=60 run

# Memory benchmark
sysbench memory --memory-total-size=100G run

# Disk I/O benchmark
fio --name=randwrite --ioengine=libaio --rw=randwrite \
    --bs=4k --numjobs=4 --size=4g --runtime=60

# Network benchmark
iperf3 -c <server-ip> -t 60 -P 4

​

Next Steps