Although this has been a popular subject for quite a while, I am not sure whether all our readers are familiar with the VDI acronym: Virtual Desktop Infrastructure. In a few words, this solution consists in centralizing workstations in order to use only a light equipment locally. The idea is to have a virtual machine (your own machine) on a server and to use it through a client that offers more services than Citrix/TSE (video, sound etc.).

After having raised the subject on several occasions internally, with other ISDs and solution providers, I can say the profitability of this type of solution is very difficult to find. Where there is a local machine, these solutions need: a thin client (an old PC can also do the job), resources on a server, a license for the server machine. With this new environment, we must find profitability when, on the one hand, we have a 600-euro PC (without OS) and, on the other hand, we have a thin client and the VDI license at 500 euros (without taking into account server resources).

The benefits that providers/prescribers often emphasize are the following:
- Less management: the thin client can be managed remotely like the virtual machine.
- Less maintenance for the equipment: The high level of reliability of today’s equipment is a thing of the past.
- Increased speed in making an environment available: Certainly one of the most important added values.

Based only on this fact, the criteria do not seem to be sufficient. The return on investment remains doubtful, especially for an offshore company like ours where the wage bill difference is not favourable here (we cannot always win)

As far as we are concerned, we have identified the following major benefits:
- Having clean environments: our development environments must remain “clean” in order to limit disruptions on execution environments and carry out the best configuration management. With models at our disposal, it is easy to always use an environment originating from its stub. Moreover, when a resource arrives (even temporarily), the benefit is significant.
- Automation: The real benefit is measured based on the automation and integration of the solution into our system. If a skillful person makes the request and the environment is immediately delivered without intervention, this responsiveness can be deemed valuable. For its most part, profitability is determined by the difficulty to automate the solution and integrate it into the information system.
- Increasing the security level: Certain people rarely use the “mobile” function of their computer, mostly when working from home. In order to avoid carrying sensitive data, which is the most delicate moment, it is possible to use the remote equipment by opening a VPN connection with their tokens. Moreover, these centralized environments are more secure as the storing benefits from the implemented redundancies (RAID, snapshot etc.).
- Anonymous work place: we haven’t reached this level, yet, as most employees are sedentary, but the fact that we can use any terminal (thick or thin), find our entire environment on it, regardless of the branch office, allows to have a significant gain in “comfort” of use.
- Purchases: Currently, the infrastructure catalog must be regularly updated, as our purchases are carried out as needs arise. In this configuration, purchases concern only thin terminals and server resources.
- A better use of resources: by concentrating execution environments, there is an obvious increase in risks, but the rate of resource use is improved. With our private/hybrid/public cloud project, this type of project allows us to render our investment more profitable. A work machine could temporarily find itself in the local cloud of another branch office if power had just failed (temporarily).
- Less management: obviously, this is not the element that will bring the greatest amount of profit. But we can evaluate the gain to a half man-year. Changing a thin client does not require special knowledge (identical rewiring). Memory upgrade is done anywhere in the resource management console. Environment update (stub) is automatically deployed at the next start.
- Energy saving: to a lesser extent, we can also make energy savings: bringing together all consumer systems (disks, powerful processors etc.) and managing in parallel the optimization of the configuration (eliminating unused equipment).

What about mobile equipment? Even though solutions are beginning to arrive on the market, we are not envisaging to include mobile equipment into this approach for the time being. The benefits will be too insignificant as compared to the difficulties to be overcome.

Key figures:
We are estimating a maintenance and implementation gain of more than half a man-year for the infrastructure team. By making a rapid calculation on energy consumptions, we are saving 5% through this sharing of resources. There is no miraculous solution that would generate a dramatic decrease in ownership costs.

Considering that such a solution cannot be implemented all of a sudden and that prerequisites cannot be neglected:
- Centralized solutions must be in place and networks must be tailored for these services.
- The change process is not to be neglected as the public concerned by the expected gains consists of developers. The physical separation from the “sacred” machine must be explained, demonstrated, prepared in order to be rendered transparent.
- Network services must be reinforced (routing, Vlan etc.) in order to generate this new traffic.
- The infrastructure department must prepare pilot projects in collaboration with the technical department in order to implement them and make a more precise estimation of gains. This cannot be done without the customer.

In my opinion, this approach cannot be avoided, for several reasons. Our changing (developer) environments determine us to centralize our infrastructure in order to ensure responsiveness, quality services and better use of our cloud. In the “Cloud” part, this approach is also interesting for making complete (development) environments available to users outside our infrastructure. “Fixed” environments will benefit from these upgrades in order to ensure overall homogeneity, but this is already another phase.

We have been following these solutions for two years. Starting concrete projects is difficult because benefits are real but limited. We must prepare to industrialize the management of these workstations in order to meet the requirements of our service cloud project. But the other triggering element would be a significant decrease in the entry level price for this type of solution.

[Episode 03] One year after our network infrastructure upgrade
[Episode 05] A feature team for website development