Virtual Desktops, Honestly Explained

If you're an IT director or CIO looking at VDI for the first time — maybe because a vendor pitched it, maybe because your remote workforce is painful to manage, maybe because your desktop refresh cycle is coming due — most of what you'll read online is written by the people who want to sell it to you. This is meant to be the piece you wish you'd read first. We've deployed over a million virtual desktops across 23 years. None of what follows is a sales pitch.

What a Virtual Desktop Actually Is

A virtual desktop is a Windows or Linux operating system running on a server in a datacenter, with the screen, keyboard, and mouse streamed over the network to the user's device. The device — a laptop, thin client, Chromebook, iPad, whatever — acts as a display, not as a computer. The operating system, applications, and data live on the server.

There are two main flavors. Persistent desktops belong to one user and remember changes between sessions — the same way a laptop does. Non-persistent desktops are cloned from a golden image at login and discarded at logoff, so every session starts clean. Non-persistent is cheaper to run and more secure; persistent is easier to support if your users need to install their own software.

The plumbing that makes this work includes a connection broker (matches users to desktops), a hypervisor (runs the VMs), storage (holds the golden image and user profiles), a profile management system (so users see their settings across non-persistent sessions), and a client application on the endpoint. That's it. Everything else is features and options.

Why Organizations Actually Deploy It

The pitch is usually "centralized management" or "work from anywhere." Those are real benefits, but they're not why most of our customers end up on VDI. The real drivers we see, in rough order:

Endpoint refresh cost. A thin client costs $250 and lasts 8 years. A business laptop costs $1,400 and lasts 4 years. If you have 800 endpoints, the math starts to matter. VDI moves the compute cost into the datacenter, where it's shared and depreciated differently.

Compliance and data containment. For hospitals, schools, and government agencies, the fact that student or patient data never physically leaves the datacenter is a simpler compliance story than chasing encrypted laptops around the country.

Onboarding and offboarding speed. A new hire gets a desktop in fifteen minutes instead of three days. A departing employee loses access at the broker in one click, regardless of whether they ever return the device.

Supporting mixed or BYOD endpoints. If your users have a mix of Macs, old Dells, iPads, and personal laptops, giving them all a Windows desktop through a browser is often easier than managing four device classes.

Remote and hybrid work. Not because VDI is the only way to support remote work — it isn't — but because VDI gives you consistent performance regardless of the endpoint and doesn't require you to ship laptops across the country.

Where VDI Disappoints People

VDI is not a cost-cutting project. Let us say that again: VDI is not a cost-cutting project. If your board is promising savings, the project is already in trouble.

A properly sized VDI stack — hypervisors, storage, brokers, licensing, bandwidth, and the ops team to run it — typically costs about the same as the physical endpoints it replaces over a five-year horizon. Sometimes slightly less, sometimes slightly more. What you're buying is not cheaper desktops. You're buying centralized management, better security posture, better business continuity, and the operational agility to reshape your desktop fleet without touching hardware.

The places VDI actively disappoints people:

Heavy graphics without GPU. Trying to run AutoCAD, Revit, Adobe Premiere, or anything that expects a real GPU on a CPU-only VDI stack is misery. You need vGPU or MxGPU and a budget for it.

Printing. Printing over a session protocol to a printer sitting next to the user has been a solved problem for about ten years, but "solved" means "works if configured correctly" not "plug and play." Budget ops time for it.

Local peripheral weirdness. USB scanners, signature pads, specialty medical devices, anything with a vendor driver that thinks it's the only software on the PC. Test every peripheral before you promise a user group will work on VDI.

Bandwidth-constrained sites. VDI is not heavy — a well-tuned session uses about 150 to 250 Kbps per user for normal office work — but latency matters more than bandwidth. A satellite link with 600ms round-trip will make users hate you, no matter how much bandwidth it has.

Users who live in video calls. Real-time video in a VDI session is a solved problem in 2022 (media redirection, Teams optimization, Zoom VDI plugins) but it's a solved problem that you have to configure. Ignore it and users will complain for good reason.

Picking a Platform Without Reading 40 Pages of Gartner

The three serious platforms for self-hosted VDI in 2022 are Omnissa Horizon (the former VMware product, now spun out), Citrix Virtual Apps and Desktops, and Azure Virtual Desktop. There is also Parallels RAS for smaller deployments, and Windows 365 for organizations that want a fixed per-user cost and no infrastructure at all.

We don't have a religious preference. We have a rough decision tree:

• If you're already a Microsoft shop with a strong Entra/Azure footprint and your users are a good fit for the cloud, start with Azure Virtual Desktop. It's the simplest on-ramp.
• If you need on-prem with GPU, strong peripheral handling, and heavy customization, Omnissa Horizon is usually the path of least resistance.
• If you have a legacy Citrix deployment and a team that already knows it, stay with Citrix. The grass is not meaningfully greener elsewhere.
• If you're under 300 users, look at Parallels RAS before committing to the enterprise platforms. It's simpler and genuinely capable.

Avoid picking a platform based on feature-parity checklists. The real differentiators are image management workflow, profile handling under load, and how your specific apps behave. Pilot the platform with your actual apps and your actual users before you sign anything.

What a Realistic Rollout Looks Like

A 1,000-seat VDI deployment done properly takes about six to nine months from contract to last user migrated. The breakdown is roughly: one month architecture and sizing, one month hardware procurement and rack, two months build and golden image work, one month pilot with 50 real users, then three to four months of phased rollout by department. Organizations that try to compress this to three months are usually the ones who call us later to unwind the mess.

Three Things to Know Before You Start

1. VDI is operational software. Budget for the ops team. One VDI specialist per 1,000 to 1,500 desktops is realistic. If you can't hire or contract that, consider DaaS instead.
2. Pilot with hostile users, not volunteers. The users who will complain loudest are the ones whose feedback matters most. Recruit them deliberately.
3. The golden image is the product. Treat it like production code — version controlled, built from a pipeline, tested before promotion. Everything else follows from getting this right.