Rethinking desktop delivery

Many IT professionals are weary of the relentless Windows upgrade spiral and the cost and disruption it brings. Rethinking desktop delivery, however, can both ease the pain and put the business onto a firmer footing for the future. Oh no, not again If there’s one type of project that’s guaranteed to get IT professionals riled …

[Anonymous Coward]

Much of the cost of an upgrade is user training and the time spent (wasted?) ensuring acceptance of the new platform. Just think of the introduction of the Office ribbon. Does it make people more productive? Hard to say, no-one knows. Was it disruptive? Oh yes.

The same problem is apparent even with browser-hosted software too - about the simplest delivery model available; so given that, virtual desktops don't really help anyone here. The pain of new software is still very much present.

We got to the stage some years back where the functionality within most office software satisfies the general use cases of pretty much everyone, so people are asking quite rightly, why the constant change, the constant costs, the constant upheavals for no real gain? And as an industry we have no good answer except the constant spectre of hackers, breaches and security problems with existing software that will all be magically solved by upgrading to version n+1. But somehow never are.

[RedneckMother]

has the worm turned?

"In practical terms, the key is to make desktops virtual. For some types of user, this could translate to total centralisation, with the operating system, applications and data all residing on the server side of the network. This approach is suitable for task-based workers who tend to sit at a desk all day running the same subset of applications."

Sound reminiscent of mainframe computing from several decades ago.

[chivo243]

99% Virtual Desktops

I'm striving for 100%, but alas, it will never happen. There are handful of mobile users that will never be virtualized, as they travel to places without any access. One can dream...

Rolling out a new VM feels a whole lot easier than upgrading desktops. If Windows 10 can deliver a solid performance in testing, then next upgrade cycle could be done and dusted easily.

At this time, I only have desktop OS's to support, but I can smell the mobile wave on the wind. Interesting times ahead.

[Terrence Bayrock]

Why not back to the Mainframes?

Hey, if 90% ( and some could argue 100 %) of your standard business applications can run on a thin client/browser type system, why bother with conventional desktop systems apps. ? load it all on the mainframe/server (cheaper to maintain). So what if the OS upgrades? Put in a new thin client/browser on a new OS (if you have to ) and off you go. Visualization can work if the virt box supports some variant of Office (since Office has the lion's share of the market).

Ducking rotten tomatoes, bad fruit and salami on the way out the door.......................

[chekri]

Hyperbole much?

FTA: "relentless Windows upgrade spiral"

Using XP as an example, it had a service life of 12 years.

To put this into perspective this, Solaris 8 had a service life of 12 years - heard any Unix admins complaining about a "Relentless Solaris upgrade spiral".

Another Reg journalism fail

[P. Lee]

Re: Hyperbole much?

>Using XP as an example, it had a service life of 12 years.

and what was Vista's lifespan, Win7's lifespan? How many people use Solaris on the desktop? This is an article about the desktop with its zillions of software permutations, not tightly controlled server OS's.

If you had the floss *nix desktops, you could break the upgrades down. Upgrade the kernel, leave the GUI for later. You can upgrade samba on its own. Upgrade CUPS before the kernel upgrade. The loose coupling means you aren't hit with a lots of problems from all different sources and all different parts of the organisation at the same time. You might have the same overall failure rates, but you get less grief from smaller maintenance changes than "we rolled out a new desktop OS and had loads of problems." I'd be surprised if that model will ever be available from MS without a subscription service. It leads to people thinking about the value of the OS package when its broken out so explicitly.

My gut feeling is that people over-centralise due to the need to squeeze value from license costs and expensive hardware. How about distributing thin clients to users and putting people's desktop machines in a room in their office building, not some expensive and remote DC? Then you could actually use your switching infrastructure to spread the load, rather than just ending up pumping all the traffic down one pipe. Dual partitions on each disk gives mean you could duplicate the working image and then upgrade it, leaving the user, and you, a fall back. Also, no massive I/O requirements on a server disk trying to do a squillion desktops. Get DNS running nicely and they could VPN into the DC network and connect out to their own desktop at the office. No expensive server hardware required, no expensive virtualisation licenses, just plenty of bandwidth to each host from each thin-client, providing decent latency figures and maybe even dual-screen functionality while in the office and guaranteed LAN-based network capacity between the management servers and desktops.

Maybe someone could put together a simple rack chassis with desktop class modules (moonshot?), not stupidly-priced server ones. AMD CPU, 8-16G RAM, 256G SSD and two GigE NICs.

[Roland6]

In practical terms, the key is to make desktops virtual

For many businesses, the bit that matters is the application. I suspect that many business would happily run Office 97 complete with it's original UI, if MS still sold it and supported it on Windows 8.

But here in lies the problem, lots of software applications have been designed to be used by a single user on a dedicated desktop. For virtual desktops to really take off we need software designed to be used by multiple users - just as it was in the days of mainframes and minicomputers. What companies such as MS are missing (and so are many Open Source projects), is the mainframe/minicomputer version of Office ie. single memory image used by many users - potential opportunity for Uniplex to reappear.

[Anonymous Coward]

Re: In practical terms, the key is to make desktops virtual

Nothing new here. I have been doing this for years deploying applications on terminal server/RDS servers. However, a large part of my job does involve getting shitty apps designed to run on a single PC to play nicely with a terminal server.