▲
AUTHOR PROFILE • SYSTEMS ARCHITECT

Pawan Mahabir (Poexone)

Creator of Poexone Labs • Maintainer of Hall 3B • Self-Hosting Purist

The Laboratory Mission

Poexone Labs was founded on a simple engineering conviction: True digital autonomy begins with sovereign infrastructure. In an era dominated by opaque cloud rentiers, recurring software subscriptions, and walled corporate platforms, we design, benchmark, and document production-grade homelab systems running on local hardware.

From clustering low-power mini-PCs and tuning Linux kernel namespaces to automating zero-trust overlay meshes and building living retro social software simulations, this platform is an open notebook of lessons learned in the trenches of self-hosted infrastructure.

THE TRIFORCE FLEET // TOPOLOGY & WORKSTATIONS

Six Nodes, Three Workstations, One Mesh

The Triforce Fleet is a hybrid mesh: three on-premise Proxmox hypervisors on a fiber uplink, three geographically spread cloud nodes, and three workstations, all stitched together by an encrypted WireGuard overlay. The goddesses guard the realm; the dragons patrol the skies.

ON-PREMISE // THE REALM

Three Proxmox VE 9 hypervisors running unprivileged Debian containers. The two mini-PCs and the network core ride on battery backup.

⚡ DIN COMPUTE & MEDIA
GODDESS OF POWER

The workhorse hypervisor. Runs the household application stack in lightweight containers: media streaming and audiobooks, password vault, documents, budgeting, GPS family map, the AI chat front-end, and the primary backup server with encrypted offsite sync.

BEELINK ME MINI • INTEL N95 (4C) • 12GB • 1TB NVMe + USB BACKUP DATASTORE
🛡️ NAYRU PERIMETER SENTRY
GODDESS OF WISDOM

Perimeter guardian. Redundant DNS sinkhole, single-sign-on reverse proxy, network discovery, home automation, and the privacy egress gateway that masks the home connection. Also the UPS commander that orchestrates graceful shutdowns.

BEELINK ME MINI • INTEL N150 (4C) • 12GB • 64GB eMMC • 2.5GbE
🌩️ FAROSH NEW • AI & STORAGE
LIGHTNING DRAGON

The newest node and the only one with a real GPU. Hosts always-on local AI inference, a household NAS, a second on-site backup mirror on spinning rust, and isolated sandbox VMs.

ALIENWARE AURORA R11 • i5-10400F (6C/12T) • 16GB • RTX 2060 6GB • 1TB NVMe + 2TB HDD
CLOUD // THE GUARDIAN DRAGONS

Lean, single-vCPU nodes in three cities: public ingress, replicated databases, monitoring, and mesh exit points.

🌿 FARORE MIDWEST SENTRY
GODDESS OF COURAGE

Midwest vantage point. Streaming database read-replica, outside-in perimeter audits, a caching DNS upstream, and a mesh exit node.

CHICAGO VPS • 1 vCPU EPYC • 1GB • 25GB NVMe
❄️ NAYDRA PRIMARY CLOUD
FROST DRAGON

Primary cloud anchor. Serves this blog, hosts the primary database cluster, uptime monitoring, the push-notification hub, and the encrypted weekly cloud backups of its own services.

SILICON VALLEY VPS • 1 vCPU EPYC • 1GB • 25GB NVMe
🔥 DINRAAL MEDIA & SOCIAL
FIRE DRAGON

Secondary cloud compute. Runs the Hall 3B living 2005 social-simulation engine, plus stream caching and indexing services.

SECONDARY VPS • 1 vCPU • 2GB • 50GB SSD
WORKSTATIONS // THE HEROES

Where the work actually gets done.

💻 POE-XPS DAILY DRIVER
COMMAND WORKSTATION

Primary mobile engineering workstation on Omarchy Linux (Arch + Hyprland). Runs the local AI swarm and a living, interactive lab dashboard as its desktop wallpaper.

DELL XPS 14 • CORE ULTRA X7 358H (16C) • ARC B390 • 32GB LPDDR5x • 1TB NVMe
🖥️ POE-LAB OMARCHY TESTBED
LAB BENCH

A 2017 Lenovo convertible rescued from Windows and reborn on Omarchy Linux. A deliberately modest testbed: if a config, script or dashboard runs smoothly on a dual-core Kaby Lake, it runs anywhere.

LENOVO IDEAPAD FLEX 4 14" 2-IN-1 • CORE i5-7200U (2C/4T) • HD 620 • 8GB • 256GB SSD
🎮 POE-LT WINDOWS & GAMING
WINDOWS FIELD RIG

The heavyweight portable. Windows 11 gaming and creative rig, and the box for anything that needs serious GPU muscle on the go.

ALIENWARE m18 R2 • CORE i9-14900HX (24C) • RTX 4080 12GB • 32GB DDR5 • 1TB NVMe • 18" QHD+ 165Hz
ᚷ EDGE PERIMETER // WI-FI 7 GATEWAY
ASUS RT-BE58U on a 2.5G fiber uplink • layer-2 IoT isolation • zero inbound port forwards
🔋 POWER & RECOVERY // 3-2-1
UPS-backed core with automated graceful shutdown & push alerts • nightly snapshots mirrored on-site & encrypted to the cloud
● ZERO WAN EXPOSURE (ENCRYPTED MESH ONLY)

Direct Correspondence

Whether you are architecting a multi-node Proxmox cluster, debugging WireGuard MTU bottlenecks, or want to discuss retro web engineering, my terminal is open:

EMAIL: [email protected]
PROJECTS: https://hall3b.com (2005 MySpace Social Sim)
LOCATION: North America

KNOWLEDGE BASE // CORE ETHOS & ARCHITECTURE

Frequently Asked Questions (FAQ)

Our foundational engineering convictions, design philosophy, and technical decisions.

Part I: Core Philosophies & Engineering Ethos

What is the core philosophy behind “Sovereign Computing”? SOVEREIGNTY

Sovereign computing is the conviction that individuals and engineers should genuinely own their digital foundation rather than renting it from corporate cloud monopolies. When your personal finances, password vaults, family records, documents, and system telemetry live on proprietary cloud servers, you are perpetually subject to arbitrary pricing hikes, terms-of-service revisions, surveillance telemetry, and vendor lock-in.

We treat our homelab as an independent, sovereign data center: self-hosted, self-managed, and fortified with air-gapped 3-2-1 cryptographic backups that ensure our personal data remains ours forever.

Why “Triforce Labs”? What is the connection between Zelda and DevOps? DESIGN SYSTEM

Enterprise infrastructure software is frequently sterile, cold, and devoid of soul. The Legend of Zelda: Ocarina of Time aesthetic is not merely aesthetic nostalgia—it is an architectural metaphor. The three Golden Goddesses map directly to the foundational requirements of high-availability distributed systems:

  • Din (Power): Compute muscle, application hosting, and the primary backup vault.
  • Nayru (Wisdom): Perimeter defense, DNS sinkholes, encrypted routing, and zero-trust authentication.
  • Farore (Courage): Geo-distributed sentries, disaster-recovery replicas, and outside-in audits.

Together they form the Triforce Fleet. Where the goddesses guard the realm, the dragons patrol the skies: Naydra, Dinraal and the newest arrival, Farosh, the lightning dragon that brings GPU-powered AI and bulk storage home.

Crafting glassmorphic HUDs, custom sound chimes, and retro-futuristic telemetry transforms daily server maintenance from a mundane chore into a sacred craft. Form and function should elevate one another.

What is the “Local-First, Edge-Sovereign” AI Principle? LOCAL AI

We believe artificial intelligence works best close to the user’s own silicon. Day-to-day DevOps work, automated file operations, log triage, and bash scripting run on local AI on our own workstations, with zero telemetry leaving our network, and a dedicated GPU node in the fleet provides always-on local inference for the whole household. We keep testing local AIs to find the setup that needs the fewest paid tokens, sending work to commercial models only when it is genuinely worth it.

Commercial cloud models are treated strictly as secondary, non-critical circuit-breakers—never as the primary brain. If our upstream fiber connection goes dark, our automated AI capabilities remain 100% functional.

What is the “Radical Simplicity” rule regarding hardware and power? EFFICIENCY

The best homelab is not the loudest one with a 42U rack consuming 1,200 watts of wall power. We prioritize silent operation, surgical resource footprints, and maximum compute-per-watt efficiency.

The always-on core runs on fanless-class Intel N-series mini-PCs with unprivileged Debian LXC micro-containers instead of bloated virtual machine fleets, so the essentials sip a few dozen watts and keep running on a modest UPS. The one power-hungry box, the GPU node, earns its keep: it is reserved for the work that genuinely needs a graphics card.

What is your “Defense-in-Depth” security posture? ZERO TRUST

We assume that every perimeter barrier will eventually be probed. Consequently, our security model never relies on a single wall:

  1. Total Port Isolation: Zero public port forwards on our residential edge router. Attackers scanning our WAN find zero entry points.
  2. Sealed WireGuard Mesh: Nodes communicate strictly over encrypted Tailscale overlays.
  3. Forward-Authentication SSO: Authelia verifies multi-factor session state at the reverse proxy layer before packets ever reach application memory.
  4. Micro-Segmented VLANs: Smart cameras and consumer IoT devices are isolated from private subnets with automated rogue DNS port redirection.
Why publish all architectural playbooks and post-mortems publicly? OPEN SOURCE

Building a dependable self-hosted architecture is daunting, often clouded by fragmented forum posts, out-of-date tutorials, and breaking changes. Poexone Labs exists to document real-world architectural lessons and authentic post-mortems in full detail—free of corporate sponsorships, affiliate spam, or biased endorsements.

If our documentation empowers a fellow engineer or enthusiast to reclaim their digital sovereignty, prevent data loss, or design a more resilient infrastructure, our mission is achieved.

Part II: Systems Architecture & Tooling Decisions

Why use Proxmox VE over bare-metal Docker or unRAID? HYPERVISOR
Proxmox VE provides true type-1 hypervisor isolation combined with lightweight Debian LXC containers. It allows running unprivileged microservices at near-zero overhead while cleanly separating compute workloads (Din) from perimeter ingress sentries (Nayru) and GPU/storage duties (Farosh). Proxmox Backup Server takes nightly deduplicated snapshots, mirrors them to a second on-site datastore, and an encrypted copy goes offsite to the cloud.
Why run a Tailscale WireGuard mesh instead of opening router ports? NETWORKING
Zero-trust overlay networking eliminates public attack surfaces. By peering nodes over WireGuard, compute servers remain completely dark to the public internet, accessible only through authenticated edge proxies or encrypted administrative nodes. Attackers scanning our residential public IP find zero open application ports.
How does Poexone Labs architect local AI with cloud resilience? RESILIENCE
We operate a 3-tier swarm protocol: Tier 1 runs on local AI over our private mesh, where we continually test local models to find the combination that keeps paid token usage to a minimum. If local compute is sleeping or exceeds a 2.5-second SLA, a fast circuit-breaker fails over to a cloud model for sub-second responses, backed by autonomous heuristic lore engines for zero downtime.
How does the blog monetize without compromising user experience? PRIVACY
We strictly enforce Google AdSense manual placement standards: Google “Auto-Ads” are permanently disabled to prevent injected floating banners or page reflows. Advertisements are confined exclusively to designated container slots, accompanied by clear disclosures, and our analytics are 100% cookieless and first-party.
← RETURN TO POEXONE LABS HOMEPAGE