Provider Documentation

Overview

VectorLay providers contribute GPU hardware to the distributed inference network. Customer workloads run in isolated VMs on your hardware, and you earn revenue based on GPU utilization.

Requirements

• GPU: NVIDIA GPU with VFIO support (RTX 3090, RTX 4090, A100, H100, H200)
• CPU: Modern x86_64 with IOMMU support (Intel VT-d or AMD-Vi)
• RAM: Minimum 32GB, recommended 64GB+
• Storage: 500GB+ NVMe SSD
• Network: Stable internet with 1 Gbps+ bandwidth
• OS: Ubuntu 22.04 LTS or 24.04 LTS

Setup Process

1. Register as a provider

   Create an account and register your organization as a provider from the dashboard.

2. Configure IOMMU and VFIO

   Enable IOMMU in BIOS and configure VFIO passthrough for your GPUs. This allows VMs to have direct, bare-metal GPU access.

3. Install the Nomad client

   Run our setup script to install and configure the Nomad client on your node. This handles job scheduling and VM lifecycle.

4. Connect to WireGuard

   Your node joins the VectorLay WireGuard VPN mesh. All traffic between nodes and the control plane is encrypted.

5. Verify and go live

   We verify your hardware meets requirements, then activate your node to start receiving workloads.

Security & Isolation

Customer workloads are fully isolated from your host system:

• Each workload runs in a QEMU VM with its own kernel
• GPUs are passed through via VFIO — no shared drivers
• IOMMU enforces memory isolation at the hardware level
• VMs have no access to host filesystem or other VMs
• All VM data is wiped when workloads terminate

Earnings

Providers earn based on GPU utilization hours. Earnings vary by GPU type and are paid out monthly. View real-time earnings and utilization metrics in your provider dashboard.