Products › VDI

Virtual Desktop Infrastructure

A co-equal focus area alongside RDP and Proxmox — currently in development, release timing tied to revenue from the foundation products

This focus area is real but the product is not yet released. Lamco-vdi is under active development in a private repo — the architecture is settled and most of the code is written. Public release follows revenue from the RDP and Proxmox focus areas, which fund VDI development.
FOCUS AREA

Wayland-native VDI for Linux infrastructure

Virtual Desktop Infrastructure is a different product class from remote desktop sharing. RDP shares an existing desktop session that’s already running. VDI creates virtual desktop sessions on demand — headless on the server, materialized only when a client connects, broker-managed, multi-user, designed for fleet operation.

Lamco’s VDI focus area is co-equal with RDP and Proxmox — not a sub-category of either, even though it builds on the same Wayland-native foundations. The anchor product, Lamco-vdi, is a headless Smithay-based compositor with a custom QUIC + AV1 transport, designed to fill the gap left by the recent collapse of independent open-source Linux VDI options.

See the RDP focus area for desktop sharing, and the Proxmox focus area for VM access — those handle different problems. VDI is the headless multi-user layer above them.

VDI vs DESKTOP SHARING

Why this isn’t the same as RDP Server

Different problem; different architecture. The product line distinction matters for procurement.

Lamco RDP Server (desktop sharing)

  • — Connects to a running Linux desktop session and shares it
  • — Single user / single session per host
  • — Requires a Wayland compositor already running
  • — RDP wire protocol — native Windows clients work
  • — H.264 / H.265 / AV1 via VA-API / NVENC
  • — Use case: remote-access an admin’s workstation, support a user, demo a desktop

Lamco-vdi (virtual desktop infrastructure)

  • Creates virtual desktops on demand — headless server, no host desktop required
  • — Multi-user, broker-managed session pool
  • — Custom Smithay-based compositor written for headless operation
  • — QUIC custom wire protocol, not RDP — thin clients use Lamco client app
  • — AV1-first encoding with budget-aware resource management
  • — Use case: enterprise VDI fleet, education labs, contractor desktops, ephemeral workstations
WHY NOW

The independent Linux-VDI market collapsed

Direct verifiable observations from the open-source / SMB Linux-VDI landscape, 2023–2026.

FlexVDI — defunct

Founded 2015, KVM-optimized VDI from Zaragoza. Ceased operations October 2023. Community confirmed dead.

Deskpool — defunct

Open-source VDI broker for Proxmox/oVirt. No longer maintained.

UDS Enterprise — expensive

~€78/concurrent user/year. Functional but heavy UI, X11-only on Linux guests.

KASM — container-only

Container-focused remote desktops. Different model than full-VM VDI.

PVE-VDIClient — SPICE-only

Proxmox-focused, locked to SPICE protocol. No RDP/QUIC, no Wayland-native.

Citrix / Omnissa / DCV — X11-only on Linux

Every commercial Linux VDI requires X11 sessions. Wayland is unsupported across the major vendors.

The combination — viable independent VDI dying off, commercial vendors stuck on X11, VMware Horizon refugees fleeing 6× license increases — defines the gap Lamco-vdi targets.

ARCHITECTURE

Lamco-vdi at a glance

Distinct from Lamco RDP Server — ~85% incompatible architecture by design.

┌─────────────────────────────────────────────────────────────────┐
│ Thin clients                                                    │
│   Lamco VDI client app  ─QUIC + AV1─▶                           │
└──────────────────────────────────┬──────────────────────────────┘
                                   │
┌──────────────────────────────────▼──────────────────────────────┐
│ Lamco-vdi server (per-host)                                     │
│   ┌────────────────────────┐    ┌────────────────────────────┐  │
│   │  Session broker        │    │  Budget-aware resource mgr │  │
│   │  • spawn / pool / kill │    │  • CPU / mem / GPU quotas  │  │
│   │  • auth, idle reaper   │    │  • Per-session encode tier │  │
│   └────────┬───────────────┘    └────────────────────────────┘  │
│            │                                                    │
│   ┌────────▼───────────────────────────────────────────────┐    │
│   │  Headless Smithay compositor (one per session)         │    │
│   │  • GLES rendering                                      │    │
│   │  • In-process app surface, no DRM/KMS                  │    │
│   │  • Capture pipeline → AV1 encoder → QUIC stream        │    │
│   └────────┬───────────────────────────────────────────────┘    │
│            │                                                    │
│   ┌────────▼─────────────────┐                                  │
│   │  D-Bus management API    │ ← shared with Lamco RDP Server   │
│   └──────────────────────────┘                                  │
└─────────────────────────────────────────────────────────────────┘

What’s shared with the rest of Lamco

  • — D-Bus management interface (common control plane with Lamco RDP Server)
  • — AV1 / VA-API encoding lessons from the RDP-server work
  • — IronRDP contributions feed back when relevant (e.g. multi-codec, autodetect)

What’s deliberately different

  • — Custom Smithay compositor (headless), not a regular desktop session
  • — QUIC custom wire protocol, not RDP
  • — AV1-first encoding pipeline; no need for H.264 backwards compatibility
  • — No dependency on the published lamco-* ecosystem crates today (codebase is standalone)
EXPLORATORY

Lamstore as a possible storage layer

Open architectural question, not a committed direction.

VDI deployments need persistent per-user state — profile data, app config, sometimes user-installed software. The default answer is “mount a filesystem,” but at fleet scale that becomes a performance and atomicity problem (concurrent profile syncs, crash-safe writes, snapshot semantics, deduplication across thousands of similar desktops).

Lamstore — a separate Lamco research project on extent-based storage with hardware-atomic page writes — is being explored as a potential persistence layer for VDI, alongside conventional filesystem and CoW options. Its primary motivation is PostgreSQL atomic-write integration (eliminating Full Page Image writes), but the same extent-store + atomic-write primitives map cleanly to per-user VDI profile storage.

What “rearchitecting it as an option” would mean

  • — Lamstore stays a standalone project with its own roadmap (PG integration first)
  • — Lamco-vdi could optionally sit on top of Lamstore for per-user persistence at fleet scale
  • — Conventional filesystem-backed deployment stays the default; Lamstore would be the high-end / large-fleet path

No commitment yet. The conversation is real, the integration shape is still open, and the storage layer is not the bottleneck for getting Lamco-vdi released. Listed here for completeness because the option exists and may inform future architecture.

STATUS & TIMELINE

Where this stands today

Repository
Private, not yet public. Lives in lamco-vdi-dev.
Code state
Architecture settled, most of the codebase written. Standalone — no dependencies on the published lamco-* crates.
Release prerequisites
Revenue from Layer 3 (Lamco RDP Server) and Layer 4 (Proxmox products) needs to be established first. VDI is Layer 5 in the company strategy — funded by the layers below it, not in parallel.
Open questions
Storage layer (Lamstore vs conventional FS), broker UI, licensing model, hosting / packaging story, Proxmox-specific integration shape.
Why this page exists today
VDI is a real focus area, not a future bullet point. The work is ongoing; the strategy depends on it. The page exists so the focus area is visibly co-equal with RDP and Proxmox — even though the product itself lands later.