On Wed, May 13, 2009 at 10:28 PM, Ward, Mike S <mw...@ssfcu.org> wrote:
> Since SLED is an Enterprise Desktop, does that mean you would have to > have one SLED for every user under VM? My approach would be indeed to run just one desktop per virtual machine, instead of what Matthew suggested with all desktops on a few Linux virtual machines. My preference would be the simplifier security issues and the ability to ensure that resources can be granted to the virtual desktop that is supposed to use them, and the ability to charge for consumed resources. Something to think about is whether the virtual machine needs to be there when the user is not. One of my pet projects was to speed up Linux boot process so that we could start the virtual machine when the first TCP/IP packet arrives (and get it done within the time that TCP/IP allows you). I even worked with a customer who considered to migrate unused virtual machines to tape and restore them when needed (and accept that it may keep the developer waiting for a few minutes). Clearly you want something to share the program code so that you can do software management in a central manner and not upgrade each virtual server separately. That requires you separate data from (centrally managed) code and server configuration. When you review the thread about "stateless Linux" on the list yesterday, it appears an attractive approach to have a small supply of "luke warm" Linux servers ready to get personalized when the user attempts to connect to the desktop. It would require their data and configuration to reside on a separate file server. I would be tempted to hibernate to disk rather than RAM (and expect z/VM paging to restore it) but either approach might work. Rob -- Rob van der Heij Velocity Software http://www.velocitysoftware.com/
