The stick, the carrot and the desktop virt project

The world would be a better place if it weren’t for all the users. Even the best laid technological plans can go awry when computer-hugging individuals decide that they don’t want to abandon their conventional systems or ways of working. Nowhere is this more true than in the nascent world of desktop virtualisation. Many users …

[Anonymous Coward]

Local Power

The problem I'm already seeing with virtualised desktops is that they're locking in at a generation behind - i.e. today we have IE9 shipping with GPU acceleration, and my MD already questions why our virtualised desktops don't look like the Aero theme he's familiar with from his local desktop, why fonts don't look as good, why Flash apps don't work as well - and can we do something about it?

Well, technically, yes, but not without giving up a lot of the cost savings of virtualisation - i.e. our virtual desktops don't have GPUs, and we use a low bandwidth model as the main thing is supporting remote workers. I know it's possible to virtualise a high-end graphics workstation, but most of the large projects I know of, tend to be aiming at a low-end spec - they are virtualising existing needs.

The big question for me is what happens as web apps become 'richer' and more dependent on a client having some grunt. And by grunt, I mean the amount of grunt available on a moderate Android phone. Will virtualised desktops be the new green-screen terminals?

[Anonymous Coward]

Arrogant report much?

Meanwhile, developers treat their machines like well, their machines, that allow them to do their jobs as efficiently as possible. That is, until they're replaced with FOTM systems like this, running a server OS with bare minimum resources and the inevitable annoying lag.

I'm sure it can be done well, but not in my experience, it's just a shiny toy for the IT control freaks. So please, developers, resist this crap, they'll demonstrate a powerful setup and tell you they can increase resources to your machines when required, then a couple of months down the line they'll actually try to take it away when their beancounter boss wants to save money, leaving you working on a laggy piece of shit with as much use as said mobile that you're so attached to.

So this developer at least, would rather hug a proper machine.

[Anonymous Coward]

A Dilbert Moment

"We have six PCs available to 500 staff in shared workrooms. If they want access to an app that won’t run in the thin-client environment, they use those and then come back to their desks."

No they won't, the good ones will leave for someplace that actually cares about productivity. This one-size-fits-all strategy might work where everyone does the same job, but reality is different. Several years ago, a large European airline got a "too good to be true" deal from Dell for thousands of desktops. Everyone got one. A 1.4 GHz P4, 256MB, 15" monitor and XP. Something that would fetch 100 quid in a boot sale. Fine for the security desk at the front of the building, HR, and the PHBs who use them only for email. A disaster for in-house developers, and anyone doing analytical work: network planning, operations research, revenue management, pricing, etc. Productivity slowed to a crawl in these important parts of the organization and IT quickly had to replace the crap boxes with something usable.

[Anonymous Coward]

Doesn't add up

We've looked into the thin client approach a couple of times, and have always found that it just does not add up. There are benefits to be had if you're starting from an unmanaged or poorly managed population of desktops, but once you have automated app delivery, automated OS installation, and decent power management software in place, moving to thin clients is likely to more power overall, be less flexible in various ways, and generally offer a poorer service. So in most cases, the problems of a poorly managed desktop environment are best solved by improving the management of it, not by migrating to VDI. If you need to keep desktops around at all, then automating apps and os delivery, etc are all worth doing, and don't cause the kind of headaches that VDI can.

On the power-saving front, a thin client generally consumes power at a flat rate, and cannot hibernate or standby when idle. A decently managed desktop, on the other hand, will consume power only when it is in use, and will standby/hibernate the rest of the time. With a modern PC, this can mean using less power overall before you even consider the power used by the extra back-end servers that the thin client needs.

The point made above about GPU acceleration is a good one. All the browsers are headed in that direction, and a thin client can't benefit from that currently. Any 'go faster by offloading stuff onto the local device' approach requires something there to offload onto...

[Anonymous Coward]

Datacenter requirements?

Although the article is about managing user expectations, this article is irresponsible. Inferring VDI will go forward more easily by managing the users without mentioning managing the CxO suite/managers about the costs and infrastructure required is irresponsible. Although VDI can cut resource requirements, you are still transferring the resource delivery from the user location to the datacenter. When the user interaction with VDI is crap because the backend resource expectations were managed incorrectly [and how many of us really have high density datacenter infrastructure?], business units will spend their operation budgets buying PCs again.