According to a Gartner report in June this year, the overall thin-client computing market will continue to grow at a steady pace during the next several years. The analyst noted that this growth will be fueled by falling thin-client terminal prices. While the thin-client approach makes Linux PCs easier to manage, there are still kinks to iron out with desktop Linux operating system, according to Gartner.
And no other OS has theses?, fact is people like symatic cannot sell their products on desktop Linux, thats the kink.
I think it’s down to money because we know that desktop Linux is as good as any other OS.
Many projects are moving more and more to thin client solutions because lot’s of older(cheap, unused) computers are idle and do nothing. One example of this movement is PC-BSD(FreeBSD)- desktop OS that can work on desktop and as server without any significant change in codebase. I think this is future of computing- we already got high-speed internet connections and for responsible terminal screen featured with 3-4Mbit of bandwidth is enough even for watching movies fullscreen with sound. Those monster computers with nearly 1kW PSU is moved into basement instead of current desktop position and start serving silent (I mean really silent), environment friendly thin client computers.
More information about PC-BSD TCR(Thin Client Release) project is found here:
http://forums.pcbsd.org/viewtopic.php?t=5028&highlight=
And what about support for multiple users in one computer? That would kick ass. People at Userful and HP already sell products based on that concept but there isn’t much support from the linux kernel project or any distribution whatsoever.
The truth is, there are quite a lot of scenarios where that concept is even more cost-effective than the thin-client solution. For example, equiping schools and multiple uses in developing countries. Moreover, serving a single computer is even easier than serving a server and it’s share of thin clients.
I don’t really understand the point of Userful. You need a second graphics card for each machine, and the software is proprietary, isn’t it? Then, you are limited in physical placement by the length of a tranditional VGA cable.
Given that an old SFF Compaq or Dell with net booting costs about the same as a graphics card, why not get one, and put your stuff wherever you want where you have an ethernet connection, which will be pretty much everywhere. Then you put your server somewhere physically secure.
Or is this missing something?
I don’t really understand the point of Userful. You need a second graphics card for each machine, and the software is proprietary, isn’t it? Then, you are limited in physical placement by the length of a tranditional VGA cable.
The advantage would be less power consumption, less noice, and less space, less generated heat.
This could be useful e.g. in school computer labs. In public environments another advanage would be that there would be less things to steel or lock to the wall to prevent theft.
If the computers are booted over the network, it will also generate less network traffic in the morning when people comes to work and boot their desktops.
There are also people who share offices, where the length of VGA and keyboard cables wouldn’t be a problem.
Besides normal office workers usually don’t need the power even from the cheepest new PC you can buy, unless they are playing games on their bosses time, so sharing computers to get better resource utilization would be a good idea.
Userful offers a very usefull (what a pun) feature which is the possibility to replace up to 10 PCs with a single one. That possibility brings great advantages like power saving, space economy, simplified equipment management, and above all, a very low cost per seat.
You mention the problem of needing a second graphics card for each machine. Well, if you were a cash-strapped organization then what would you preffer: spend 2×500$ on two PCs or spend 500$ for a single PC plus 25$ for a graphics card and 100$ for the monitor, mouse and keyboard? And the cost goes down from here. You only need to compare f(n)=500$xn with g(n)=500$+(25$+100$)xn and see what kind of money can be saved on using a multiple-user system. We are talking spending about 125$ for each seat, which is a bargain.
Edited 2006-09-25 19:03
I would spend $50 for an ex corporate Compaq. Take out the hard drive. Boot over the net from a decent server. Put the machines wherever you want, move them whenever you want. Put my server someplace secure (ie not next to the workstations). If you are going thin client, why spend $500 on a station? Its absurd. Surely, this is not the choice.
This is why LTSP was invented. But you can do better, with eg freenx. And doesn’t the Userful software cost quite a bit?
How much is the server? The server and networking upgrades to provider boot images and remote applications are not trivial. There is some consolidation: a dozen users can use a single server but server hardware tends to be more expensive than desktops.
How much work does it take to manage the server? Setting up network booting and applications for remote use is doable but not trivial. It is possible to automate updating and managing all of the desktops.
Why not leave the hard drives in the machine? Booting off the hard drive is faster than booting over the network. And is less sensitive to network capacity. Instead of running the applications remotely, run them over the network. Portability can be accomplished by authenticating to an LDAP directory and storing home directories on remote storage. This is also useful for laptops and developer machines that can’t be centrally managed.
The server is the same in either case, no? In the userful case you have the server plus multiple graphics cards and screens, and you connect up the screens over vga cables. You shouldn’t need anything special in the server except dual ethernet ports.
Yes, local booting is certainly an option.
The issue seems to be how you run off the server. In the terminal server case, you connect over ethernet and run the apps remotely on obsolete hardware at the station. In the userful case, you are connecting directly. The standard objection before freenx would have been that its too slow over ethernet, but this seems no longer to be true.
In addition, in the userful case, you have to buy the software. Two seats are in the free version. How much do you have to spend to get to half a dozen?
Debian edu would be the natural one to try. It would be interesting to hear from anyone who has tried it. I’m probably going to be having a go this winter. The server and client requirements are minimal. There’s a lot of info on http://wiki.debian.org/DebianEdu
Edited 2006-09-26 11:05
I mentioned 50$ for the graphics card because that’s the going price for brand new low-end NVidia/ATI cards. If you are going used then there are a whole lot of cards which do the job perfectly and are a dime a dozen. So, in fact, the price per seat on a multi-seat station is well below that.
So as it is easy to see the multi-seat station is the best possible alternative for those organizations which have very thin budgets but need to equip themselves.
Well, this isn’t hard to do with a Standard-PC and Linux. Some months ago there was an article about that in the German IT-magazine c’t (c’t 10/06, Page 228).
You just need multiple graphic ports (p.e. 1 card with 2 connectors) 2 usb keyboards and 2 mice and a patched X-Server. Only problem was that the sound for all users came through 1 soundcard.
In the c’t they made an setup with 4 workingplaces per PC.
thin was in, then it went out….now occasionally someone mentions it coming back in again…
Maybe to a limited extent where it can truly be beneficial but that is about it. I think systems are so inexpensive that the cost isue is not a issue.
Maybe to a limited extent where it can truly be beneficial but that is about it. I think systems are so inexpensive that the cost isue is not a issue.
No, it’s not only about the hardware anymore. There’s a lot of value for enterprises collapsing rich-clients into a datacenter / thin-client infrastructure.
A substantial driver is security and compliance requirements. Organizations want to control their data, they want it centralized.
Remote offices can hang thin-clients off a seperate data center via WAN connection, simplifying things for IT people.
Thin client with SSL gives employees secure remote access to corporate apps and data from virtually any platform without risking data leaving the network.
Client OSes are much less likely to break when not innundated with a variety of local application updates and configuration changes, making them much simpler to manage and troubleshoot when something goes wrong.
Certainly many of these advantage/requirements can be applied to conventional client/server based networks, but they often require extensive third-party products and infrastructures which increases cost and complexity.
Thin client is in. It’s not going to replace rich client across the board any time soon, probably not ever, but it’s not viewed with the same disdain anymore by CIO’s either, not when it’s both a cost-saver and a compliance tool. Both Citrix and Microsoft are doing big business in terminal / thin-client services, and Microsoft acknowledged the trend enough to release the thin-client version of XP in order to dissuade enterprise customers with little incentive to upgrade to Vista from exploring linux thin-client opportunities.
Thin client is one of the single best opportunities for desktop linux to breakthrough in corporate environments. My own company recently migrated to Citrix for all of our core apps and I can use them very easily with Citrix on my linux laptop.
Having said all that, I hate Citrix. No matter how well tuned the QoS settings are, latency always becomes an issue at inopportune times on our WAN connection. It’s awful from a useability point of view for those of us stuck in remote offices, not as much of a problem for those using LAN connections to the Citrix farm.
It’s also worth pointing out that our company went ahead with the citrix migration despite the astounding amount negative feedback from the various offices around the world that were part of the trial. Our executives justified it by claiming the advantages were significant enough that the employees would have to adapt. Probably the same reasoning that keeps us stuck on Notes as our global email platform…
Now, if only I could actually read the article, since the server seems to have crapped out.
This was just becoming evident at VMworld 2005. Moreso this year. Many companies are going to the thin client to connect to citrix or virtual desktops. And since linux runs on such lean hardware and over a network connection, it is one of the primary thin client solutions. Linux may end up gaining huge market share, but it won’t be like you think. It will be as the OS of choice for thin clients and other appliances. I see this taking over at home before long. This would fix all of the worm/security issues that are caused by users at home who dont know how to use a computer. Think about it. Your OS is downloaded from your ISP, which is an always on connection, at night. You are always updated and secure. If something goes wrong, they can send you a new OS ina matter of minutes. all that will be on your local HDD is your data. MS has already indicated it plans on going to a subscription based model for software. How else do you do that except by requiring people to keep paying for service. There will still be those of us that have regular computers, but you will start seeing the average home user going the appliance route before long.
As for HP, the reason for the 4 user machine is simple. Cost. They calculated it was less expensive to add a couple video cards, which can be multi-headed, and a few sound cards, than to buy Big Iron to run terminal services on. And while the software HP and userful sell is probably proprietary, it can be done without those tools. Just really have to know what you are doing.
er, I mean thin client.
I’ve been hearing “thin is in” every year, with pretty much the same economic and control arguments, since 1985.
As long as there have been thin clients, the first of which was probably the IBM 3270 terminal, there have been a few uses for them, but there has never been a time when the market was all that large, and there’s not one in the forseeable future.
then on second thoughts I read the article. Then I knew my first impression was right.