A developer opens their laptop on a train, connects to their workstation hosted 300 km away in a data center, and finds their files, applications, and complete environment. No data is transferred to the local disk. This is the very principle of VDI, or Virtual Desktop Infrastructure, an architecture that centralizes desktops on remote servers and delivers them on demand to any terminal.
Hypervisor and connection broker: the technical foundation of a VDI
VDI is often referred to as a simple “remote desktop,” but the mechanism relies on two distinct software components. The first is the hypervisor: it divides a physical server into several virtual machines, each running its own operating system and applications.
The second component is the connection broker. When a user authenticates, the broker assigns them an available virtual desktop from the pool corresponding to their profile. This routing is transparent to the user, but it determines the fluidity of the service.
To delve deeper into the definition of VDI in computing, it is noted that all graphical rendering occurs on the server side. The local terminal, whether it’s an aging laptop, a tablet, or a thin client, simply receives a display stream and sends back keyboard and mouse inputs.
In practice, the quality of the experience depends as much on network latency as on server power. On an unstable WAN link, a simple spreadsheet can become cumbersome to use. Network sizing is crucial for the success of a VDI project far more than the choice of the vendor.
Persistent or non-persistent VDI: choosing based on user profile
There are two deployment models, and the choice between them has a direct impact on storage costs and administrative workload.
Persistent desktop
Each user retrieves their personal virtual machine, with their shortcuts, files, and settings. This model is suitable for developers, financial analysts, or anyone who heavily customizes their environment.
The downside: each persistent VM consumes dedicated disk space. Multiply that by several hundred users, and storage becomes the primary cost factor.
Non-persistent desktop
The user receives a “clean” desktop at each login, generated from a reference image (often called a golden image). Customizations are limited or managed through roaming profiles.
This model is suitable for call center environments, rotating desks, or training rooms. Administration is lighter, updates are made on the single image, and storage remains controlled.
- Developer or analyst profile: persistent VDI, to maintain a tailored environment between sessions
- Operator or support agent profile: non-persistent VDI, with standardized golden image and roaming profiles
- Mixed profile (consultants, external contractors): non-persistent with folder redirection to shared storage for accessing project documents
Return to the office and sovereignty: what hinders VDI projects in 2025
Articles on VDI often present remote work as the main driver of adoption. The reality on the ground has become more nuanced in recent months.
Since 2024, several large companies have tightened their return-to-office policies. This movement slows down large-scale VDI projects and pushes towards hybrid architectures: VDI is retained for some nomadic profiles or external contractors, while the rest of the employees return to traditional physical workstations.
At the same time, issues of digital sovereignty weigh on technical choices. In Europe, regulations regarding data localization lead some organizations to favor VDI hosted on national infrastructures rather than DaaS (Desktop as a Service) operated by an American hyperscaler. Opinions vary on this point: some IT departments believe that sovereign DaaS is still too immature, while others feel that the gap is closing quickly.
VDI or DaaS: selection criteria for an IT department
The confusion between VDI and DaaS arises in almost every request for proposal. VDI refers to an infrastructure that the company operates itself (or via a managed service provider) on its own servers or in colocation. DaaS externalizes this infrastructure to a cloud provider.
- Control: VDI offers total control over hardware, updates, and security rules. DaaS delegates this responsibility to the provider
- Entry cost: VDI requires an initial investment in servers, hypervisor licenses, and storage. DaaS operates on a monthly subscription per user
- Elasticity: DaaS better absorbs seasonal peaks (temporary workers, short projects) without over-sizing the permanent infrastructure
- Compliance: for regulated sectors (healthcare, defense, finance), on-premise VDI often remains the only option validated by auditors
In practice, we see more and more mixed architectures: an on-premise VDI foundation for permanent workstations, complemented by DaaS for occasional needs or remote sites without server rooms.
Security of virtual workstations: what centralization concretely changes
Centralizing desktops on a server changes the attack surface. Data no longer resides on the user’s terminal. A laptop stolen at an airport contains no work files, only a connection client.
Security patches are deployed once on the reference image, then propagate to all virtual workstations upon reboot. Compare this with managing hundreds of physical machines scattered across multiple sites, and the operational gain becomes tangible.
The downside is the concentration of risk. If the hypervisor or connection broker is compromised, the attacker potentially gains access to all desktops. Network segmentation between VM pools, encryption of display streams, and multi-factor authentication on the broker are not options: they are prerequisites.
VDI does not eliminate security risks. It shifts them to the central infrastructure, where a specialized team can address them more uniformly than chasing after scattered physical workstations. It is an architectural trade-off, not a miracle solution.