Choosing the Right Client Computing Architectures for Your Business

Federica Troni talks about the changes happening on client computing, even faster in the last 3 or 4 years: we are working differently (collaboration, flexibility), the boundaries are blurring between work and personal life and there is a tendancy to re-centralize for better manageability, security.

The point of the presentation is about the 30% of the IT budget spent on sourcing client devices.

User requests are changing and the enterprise still must manage the PCs. The PCs are now seen as tactical (vs. strategic) assets. There is opportunity for optimization.

Today, there are options: virtualization, at every level (decoupling between hardware and software or between applications when they are packaged in a VM), is getting mainstream.

Server-based computing: access to an application running on the server through a display protocol (e.g. RDP). It leads to great economies of scale (50 to 80 users per server), it is scalable, mature, it has a large ecosystem of vendors BUT it is not for every user and application, it may even reduce productivity.

Blade PCs (it's a niche solution): the PC is in a cabinet dedicated to 1 or - better - a few users, accessed from a remote terminal. It provides good manageability, is used for vertical solutions mostly BUT the TCO is not so good (the capital investment remains high) and solutions are proprietary.

Hosted virtual desktop (hottest now!): servers hosting one VM for each user (with an average ratio of about 4 to 5 users per core), accessed through RDP. It allows to use standard software and gives users their usual experience, in total independance. Third-party tools are still evolving in this ecosystem.

Not a client technology by itself is the brokering software: sitting between the user and the server, its job is to "recognize" and connect the user to the appropriate environment. There are 2 types of brokers: simple "connectors" or "gateways" providing more services. Tools may evolve to provide a "wise" choice about application distribution (selecting the most appropriate model for every application: for example, hosted office applications and local AutoCad).

Shared desktop (intriguing): trying to use the idle cycles on desktop PCs (and increase CPU utilization) by sharing one between multiple users. But the Microsoft licensing model is not favourable to this model. Ncomputing has sold more than one million seats of their offering on this model (mostly in education market).

OS and application streaming: distribute software on demand, piece by piece. Applications are only "cached" (not installed) and generally virtualized on the client, eliminating incompatibility risks. One avantage is in offline use. BUT scalability is not proven, the market may be unstable and cost tend to go high. Vendors: MS & Softricity, Citrix & Ardence…

How to choose: Federica presents a table with the most important criteria. Usually 2 or 3 solutions are required to cover all needs, one size does not fit all. Some of the solutions (i.e. streaming or web-based client computing) take advantage of the power of the PC.

One central question is how much control you want to keep on users or how much autonomy you want to give them.

TCO is more than capital costs: a comparison between PCs and virtualized PCs shows the difference is not tremendous (most of the savings come from end-user costs), at about 10-15% savings.

Server-based computing appears to be the most attractive (on TCO) but it is true only if all the applications can be delivered on the server.

Beware of the confusion and mismarketing on virtualization.

Recommendations:

• Put the user requirement at the center, segment the population
• Centralization is "in" again
• Obstacles remain (immaturity, licensig implications)