Making Desktop Enterprise easier to manage.
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 It feels way too outdated.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.
LTSP
Virtualized Desktop environment
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. |
LTSP
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:
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:
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
|
All times are GMT -5. The time now is 11:04 AM. |