Out With the Old, in With the New: Choosing Practical Alternatives to the PC

Brian Gammage talks about what is happening in client computing.


Not much has happened in 2009 but there will have to be some hardware replacements and it is getting more difficult. It goes from the PC and thin client that are common today to virtualized OS, downloadable workspaces, new ways to deliver applications... There is choice! And you need a plan, otherwise it will be the supplier's plan.

The arrival of Vista has changed the mindset. The marketplace has shifted to a desire to take control.

Drivers

PC were too costly (running them) because the OS got complex. Virtualization tries to make things simpler again by decoupling layers on the desktop. Standardization and automation are the keys to lower TCO, that's what virtualization brings (run a standard image on different hardware or OSes).

OS migration is a fact of life. Then you may be migrating to Windows 7. It's an opportunity to check the new options and assess the changes that are coming. What you bring in should fit in your environment (tools, processes...).

Options

OS and application streaming: the applications lives on the server but runs on the device. Some tools are quite evolved (cache...). OS streaming is less used, because the OS is a big chunk to download, so it is viable only if the network is up to the par.

Server based computing: the applications are run from a server OS. Notebook users are a problem (when they are not connected). Some applications don't work very well (network limitations). It is well tested and works very well.

Blade PCs: the PC is stuck in the data center and deliver the front-end on a thin client. It is being used in investment banks (on trade floors), but not much anywhere else. It improves security. It is very expensive (dedicated and specific hardware).

Hosted virtual desktops: move the OS "bubbles" to the data center. It is popular but it also brings challenges on the network. It will improve (e.g. Citrix) and works for a broad range of applications.

The "pooled" HVD (deduplicate some components, such as the OS, and rebuild the image on demand): it reduces storage costs (60%), "Windows rot". The issue is that it does not allow for personalization. The next step is "persistent personalization". It works by dividing the upper "bubble" in 2 between the applications and the personalization. It broadens the target audience (some of the users who could not use HVD because of specific requirements).

Portable workspaces: carry the "bubble" (with an OS or not) not the hardware. 1.2 million seats by now and it is growing. Lack of standard and security issues remain.

Brokering software: connect users with centralized resources, with different management layers (provisioning, security...).

It all goes to new "composite workspaces", with different ways of building images. And PC configuration will change along.

How to evaluate and select the options?

There are new risks: lock-in in someone else's roadmap (VM formats, hypervisors, management tools...).

The initial problem was the complexity of the management: don't forget that the objective is to reduce it. You need to offset the new tasks with reductions on others.

There are changing boundaries on responsibilities (data center vs. desktop helpdesk). Education is needed.

Think how to deploy things to users. Adjust the models to the user requirements, keeping the future in perspective. There should be 3 to 5 different user profiles.