I'm sorry for the long post. I want to hear your opinions. There are many Desktop solutions out there. XP is near death. This recession is making Sysadmins to think creative and come out with the easier and cheaper solution for the future growth. The physical server and client hardware concept seems so 1980s to me now. It feels way too outdated. There are way too many drivers, chipsets, hardwares, distros, and way to many incompatible layers among each OS releases for all various Oses. The needs and demand for corporate apps for desktop haven't changed much over the years, but the user Oses have became bigger and hardware lists are getting larger. We are at the point of IT costs are way too much for most companies to manage. Most of the IT costs are dedicated to the end users. Many legacy corporate XP apps no longer work on Win7 now. This means there will be many rewrites with corporate applications. For most companies, this is the best time to rethink their migration strategies. Every year the cost of maintenance seems to go up thanks to Microsoft. Enterprise IT wants their user environments to be simple. The simpler the solution for the end users, it lowers the overall cost of maintenance. It seems like a right time to slowly wipe clean from all the useless legacy ideas and move on to the next level. Easier management and clean usability for the end users always have been the key concern for the enterprise. If it isn't easy for the end users, nothing can be sold. One thing for certain, servers will go virtualized. It makes sense to keep servers contain in virtualized containers for easy migration, redundancy, deployments, and lower energy costs. Long as users can't tell the changes, it doesn't matter what gets deployed in the NOC. However, there seem to be many options out there for Desktop or end user node strategies. I think the concept of a full blown desktop OS is becoming less desirable. Linux desktop is wonderful, but the driver issue always has been the greatest battle. Enterprise can't afford to spend more than one hour for a technician to fix the driver related issue on the desktop. This process is non-effective. Since, the solution is nowhere near, deploying the physical desktop OS without the virutalization isn't a solution. Virtualized desktop creates self contained driver layers, which reduces the hardware troubleshooting difficulties in the future. It simply saves money if the management for the end users gets easier.

It seems like there are few Desktop solutions.

Back to the late 90s with thin clients with web apps, RDP, and NX.

1. thin client solution makes outsourcing and hosting applications easier than ever. Creating a redundant network is easy. No need to spend incredibly amount of money hiring network consultants to build redundant internal networks. If the security of the apps are tight, it is very easy to offload the apps to the cloud in case of hardware failures. The future corporate database will be ran off the servers without no need to tweak the client configurations. Corporate applications shouldn't depend on the local client libraries. Future corporate apps will be php/javascript/mysql that can be easily ran and contained from the server.
2. Bandwidth costs are getting cheaper and more reliable.
   • No need to assign a dedicated router and laptop for remote users.
   • Just give them a remote connection app on a USB stick that is self contained from the OS.
   • You can even give them a self-contained virtualized image that already has VPN connections, networking maps, and other network configurations.
3. Better, cheaper, and more reliable networking hardware exist now than five to ten years ago.
   • All the difficulties of making a thin client solution viable in the past have disappeared.
   • Designing an embedded system is easier than ever with various hardware options and various open source solutions.

LTSP

1. It gets rid of the entire concept of your local disk. PXE network boot is definitely viable to even mid to large size corporations. Thanks to gigabit ethernet to the desktop, cheaper ram cost, cheap and powerful servers. Deploying LTSP is easy and users will not know the functional differences.
2. Never have to fix the client node.
   • If something breaks, swap it out with another thin client.
   • When I read the help desk logs, most of the troubleshooting issues are based around changes in the users desktop environment and false positive virus and malware issues. Most of them aren't critical changes that were made to the servers. Thin client makes these issues disappear.

Virtualized Desktop environment

1. Driver and hardware headaches of managing desktops completely disappear. The biggest headache of IT department are manging driver issues, deploying patches, virus and malware attacks, and managing desktop images.
2. Virtualized Desktop environment means it is easy to write few cron jobs to boot the virtualized desktop and run an automatic maintenance.
3. Most newer thin clients support VDI protocol.

Now, I have few questions. Which solution have you rolled out? What difficulties have you faced? What are the pros and cons of each solution? What are cost issues have you faced?

I have one more question. This always have been a difficult one to solve. I have a question about WINE with OFFICE. I don't see the future in Microsoft Office suite. It is expensive and very time consuming old idea. Deploying MSOffice service packs cause more headaches than installing a new revision. However, the users need the temporarily patching. You can't swap things out in one sweep movement. Most companies already have licenses for the legacy Office 2003. Running Office 2003 on WINE seems very stable and writing a simple registry script to change the license number through a registry editor is very easy. It isn't very difficult to run MSOFFICE 2003 from Linux servers. However, assigning an entire wine folder for each users seem to be very resource hungry to RAM, storage, and traffic. Is there any solution for users to share a WINE folder among various users? If I give them a symbolic link of the shared wine folder, how would each users maintain their own configurations?

I believe the thin client with virtualized Linux servers have became very viable solution.
Long as printers are assigned, network shares are mounted at boot, usage of file mangers are easy to use, and they have a start menu at the bottom left corner, users seem to only care about the apps running on the client node. Long as they don't have to maintain any administrations, users don't care about what Oses they are running. They only seem to care about the apps. As we move more apps to web interface applications, the old MS client and server strategies seem very outdated.

I use LTSP for small operations, 1 to 6 servers. It is a beautiful solution where IT support is minimal. Instead of having to hover over a hundred or more PCs, I get to hover over a single server. Typically, I have a few identical terminal servers that can be maintained by a simple script and ssh. I have no idea what the upper limit of this technology is but it seems that using thin clients increases the number of seats one sysadmin can handle by an order of magnitude. It certainly is no more difficult to manage 30 users or 300 as far as maintaining machines. Resetting passwords is another matter ...

Recent numbers on thin clients:

• about 10% of PCs are thin clients these days and the number grows rapidly
• China has huge over-capacity in thin clients and prices are less than $50 for some in quantity
• back-of-the-LCD-monitor thin clients cut PC footprint in half
• power consumption as low as five watts make the monitor the main load
• shared memory in GNU/Linux systems permits twice as many seats on a given server than that other OS 8-)

I would not go back to using thick clients even with GNU/Linux except for full-screen video or other heavy loads. There is no need and thin clients with GNU/Linux give better performance than thick clients:

• shared memory means almost everything is preloaded
• RAID on the server means disc I/O is superior
• services on the terminal server suffer virtually 0 network lag
• GNU/Linux is faster than that other OS anyway

Even for heavy loads, if the load is big enough one should work on a server or server farm. The thick client is rarely necessary these days.

Our network is going to gigabit/s this year so things will only get better for GNU/Linux on terminal servers.

Linux community is missing a big important tool. It is missing business papers and documentation projects. There are plenty of white papers written by Novell and IBM, but they always been useless piece of junks. They are long, boring, and primary focus on their point of views. Most busy Linux/UNIX sysadmins are way too busy to re-translate complex business documents for the upper management. I wish I had free time to start a Linux business documenting project. What Linux community is missing are proposal documentations. Only few good technologists can wear both business and technology hats. Linux gurus are lousy sales people. Linux is ready in few aspects of deployments. We can't wait for vendors get on board. Hardware vendors will still focus writing drivers for Windows, because that is their bottom line. Many will not get on board until Linux Desktop market reaches over 10%. Do we wait that long for our hardwares to be always plug and play? Linux has a superior plug and play action than Windows, but many users assume Windows has a better plug and play, because more vendors write drivers for it. 90% of Windows drivers are written by vendors. It isn't Microsoft. In order to beat Windows, we need more drivers in our kernel. However, like I said it before, we shouldn't wait. I am slowly finding out. In order to win, we really need to sell new infrastructure ideas. Windows has become a fat big pig who can't keep his weight down. Bigger the weight more expensive the cost for the IT department. We need to focus on the thin client ideas. We all know well planned Linux hardware migration means there will be no typing in the terminal