Installation Guide

Copyright (C) 2026 Dyber, Inc.

Building from Source

Prerequisites

Tool	Version	Purpose
Rust	1.75+ stable	Build all Zernel crates
Python	3.10+	ML framework stack
Git	2.x+	Version control, experiment tracking

Linux-only (for full BPF features): | Tool | Version | Purpose | |------|---------|---------| | clang/llvm | 14+ | BPF program compilation | | libelf-dev | -- | ELF parsing for BPF | | linux-headers | 6.12+ | Kernel headers for BPF | | bpftool | -- | BPF program management |

Build Steps

# Install Rust
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
source ~/.cargo/env

# Clone the repository
git clone https://github.com/dyber-pqc/Zernel.git
cd Zernel

# Build all components (release mode)
cargo build --workspace --release

# Verify the build
cargo test --workspace

Install the CLI

# Install the zernel binary to ~/.cargo/bin/
cargo install --path zernel-cli

# Verify
zernel --version
zernel doctor

Install zerneld (Observability Daemon)

# Copy binary to system path
sudo cp target/release/zernel-ebpf /usr/local/bin/zerneld

# Start in simulation mode (no BPF probes)
zerneld --simulate

# Or with BPF probes (requires Linux + root)
sudo zerneld

Install the Scheduler

# Copy binary
sudo cp target/release/zernel-scheduler /usr/local/bin/zernel-scheduler

# Create config directory
sudo mkdir -p /etc/zernel

# Generate default config
zernel-scheduler --dump-config | sudo tee /etc/zernel/scheduler.toml

# Start (requires Linux 6.12+ with BPF feature)
sudo zernel-scheduler

Cross-Platform Development

BPF compilation is gated behind a Cargo feature flag. On macOS and Windows, the crates build and test in userspace-only mode:

# Works on macOS, Windows, Linux
cargo build --workspace
cargo test --workspace

# Linux-only: build with BPF support
cargo build --workspace --features bpf

Hardware Requirements

Minimum (Development)

Any x86_64 CPU
8 GB RAM
Any NVIDIA GPU (for testing CUDA detection)

Recommended (Production)

Component	Specification
CPU	AMD EPYC or Intel Xeon (dual socket for NUMA testing)
RAM	256 GB+
GPU	NVIDIA A100 80GB SXM or H100 NVLink
Network	Mellanox InfiniBand HDR or 100GbE RoCE
Storage	NVMe SSD (for dataset I/O)

Tested Hardware

Hardware	Status
NVIDIA A100 80GB SXM	Full test suite
NVIDIA H100 NVLink	Full test suite
NVIDIA RTX 4090	Basic tests
AMD EPYC (dual socket)	NUMA tests

Zernel Distro Installation

Note: The full Zernel ISO is under development (Phase 4). For now, use the component-level installation above on an existing Linux system.

When available, the ISO will be installable via:

# Boot from Zernel ISO, then:
zernel-install

# Or unattended:
zernel-install --config zernel-install-unattended.yaml

The installer will automatically: 1. Partition disks 2. Install the Zernel kernel with ML-optimized config 3. Install NVIDIA drivers + CUDA toolkit 4. Install zernel-scheduler, zerneld, and zernel CLI 5. Apply sysctl tuning 6. Install the ML framework stack (PyTorch, JAX, vLLM, etc.)