Thin Client, Or Fat Client? That Is the Question

theodp writes "If virtual desktops are so great, asks Jonathan Eunice, then why isn't everyone using them? However encouraged folks are by the progress virtual desktop infrastructure (VDI) has made, and however enthused they may be about extending the wins of server virtualization over into the desktop realm, you don't see analysts and developers eating the virtual desktop dog food. And even the folks you meet from Citrix, Microsoft, Quest, VMware, and Wyse — the people selling VDI — use traditional 'fat' notebooks. So, are you using virtual desktops? Why, or why not?" I wonder how long the abbreviation VDI will stick around.

No, he's not

[zn0k]

> If VDI (virtual desktop infrastructure) is so great, then why aren't you using it?

Eunice isn't saying that, he's quoting Brian Madden as saying so and then gives his opinion on why he thinks they sooner or later will.

You can tell because of the sentence directly before the one quoted above:

>Virtualization analyst Brian Madden asks an excellent question:

But hey, fuck accurate summaries, right?

Anyone who asks this question should not be in IT

[RobertM1968]

Really, this is so simple. There aren't suitable thin client options for most businesses. And Microsoft is to blame. No, this isn't an anti-Microsoft rant - there's no reason for them to support (or more accurately, PUSH) such a model. It boils down to this: Windows is hardly "thin client" status anymore. Computers are dirt cheap. Buying a thin client machine costs about as much as buying the cheapo level desktop most businesses need (the ones that need more powerful hardware aren't suited for thin clients, eg: CGI and video editing). Microsoft used to (still may?) charge the same license fee for the thin client as they would if it was a full fledged desktop and full OS.

Thus, what's the purpose of spending the same amount of money for a thin client machine that one would for a full fledged desktop and full OS?

It doesn't matter how wonderful the technology behind thin clients is, or how wonderful it gets... it's a waste of money for most scenarios.

And of course, Microsoft's business model is in better shape without thin clients... more support people, more certifications, more money generated. Smarter business approach for them.

It's the connectivity

[Todd Knarr]

It's all a matter of connectivity. If you're using a traditional "fat" desktop (or notebook), you're self-contained. All your software's there, you aren't dependent on any connectivity to the outside world to get your work done. A "thin" virtual desktop client, by comparison, is completely dependent on having a network connection to it's host server to operate. Without that connectivity, it's a doorstop (and a light-weight one at that, so it doesn't even do very good at blocking a door open). And in a world of corporate firewalls and filters there may not be any connectivity that the VDI client can use. Anything other than HTTP/HTTPS may be blocked completely, and HTTP/HTTPS traffic will usually be forced through a proxy server that, even if it allows the kind of streaming connection a VDI client needs, introduces so much delay that the desktop becomes useless. And that's when the network's working correctly. Add in random network outages and traffic congestion at the wrong times and corporate systems that require non-corporate machines to VPN to the corporate network (and to have specific anti-virus and management software installed before the VPN's allowed to connect) and it makes a VDI client distinctly unreliable and hard to deal with. Meanwhile, the guy with the "fat" notebook may have more system management headaches and software synchronization issues than the VDI system, but he's still getting his work done while the VDI guy's sitting twiddling his thumbs while the techs try to sort out all the problems.

Re:Performance

[cratermoon]

The licensing costs end up being the key issue in companies of any size. By the time they set up and license all their people with client machines and all the applications, a company will spend about as much as just buying PCs in bulk from Dell or whoever and site licensing the corporate-standard MS Office suite. Pile on top of that the various fiddly things about virtual desktops that just don't work like having a real desktop PC raising the support costs and it's not competitive.

The central server with dumb terminals era ended long ago, except in niche applications. Desktops and laptops that a capable enough are just too cheap and standardized desktop support contracts from third-party support operations pretty much rule the budget considerations. For virtual and really thin clients to take off, the licensing would have to be notably cheaper and support for the edge cases like traveling remote access would have to be much better.

Re:Performance

[Mordok-DestroyerOfWo]

Another problem is buying the thin clients themselves. I can't remember off the top of my head, but the price difference between a client and a new workstation is barely enough to offset the difference in licensing, not to mention user complaints on performance. We toyed with the idea of retrofitting older workstations but supposedly we'd miss out on VMWare's/Teradici's proprietary PCoIP protocol. When it comes down to it, it's cheaper for us to buy more RAM and replace parts as needed on our existing workstations than purchase the needed infrastructure and thin clients to even begin replacing our workstations. I have begun giving some of our secretaries (or whatever the PC term is this week) older workstations with Debian with few complaints and more than one compliment on performance, however our GIS and admins will always need Windows and a beefy graphics card.

Um... "Virtual Desktops"?

[Admiral Burrito]

I've been using vitual desktops [wikipedia.org] since FVWM in the mid 90s, and it has nothing to do with what this guy is talking about. I'd think Slashdot would know better, but of course times have changed. Am I going to have to start calling it Spaces [wikipedia.org] now?

Re:hmm

[hawguy]

Why not just use a server? Amazon EC2 will rent you a "micro instance" Win2003/2008 server for 3 cents/hour ($21.60/month) or an Ubuntu server for 2 cents/hour ($14.40/month) plus a few bucks for storage.

http://aws.amazon.com/ec2/pricing/ [amazon.com]

Just don't do anything I/O or bandwidth intensive since you also pay for I/O and bandwidth.

It may be old school but it works

[emeitner]

At my place of employment, 250 employee co-operative retail with three locations, I set up a 2 node DRBD/Heartbeat cluster. It is running NFS, Samba, LDAP. Clients, 42 of them, g are $275 Zotacs(Mag HD-ND01-U) running Ubuntu 10.04. I developed a disk image with everything the way we want it. It takes me 10 minutes to set up a new machine and most of that is the unboxing part. Clients authenticate via LDAP and mount NFS homes via autofs. Some apps are local such as Firefox and Thunderbird. Other business apps are accessed via A XenApp/Citrix server using the Citrix Native Linux client. And then there are the HR and Finance SAAS applications. Now the clients could just offer a RDP connection application and the Citrix server could be a server providing virtual desktops. But why? It would add a few more layers of complexity with little benefit. The client machines are cheap, fast, easy to replace. The OS is free. The user gets the performance of silicon on the desk with the storage reliability of a server in the closet.

Re:Performance

[Daniel Phillips]

The licensing costs end up being the key issue in companies of any size.

Let me see, the licensing costs for all software I use at work is, um, $zero. And the licensing costs for all software I use at home with the exception of games is, um, also $zero. And I use a lot of different kinds of software. That's because everything I need is available for free on Linux, at home and at work. If that is not the case for you well you have my sympathy or maybe it's just your own darn fault.