[virtmach] Implementing VMs efficiently

Sean McDirmid mcdirmid@cs.utah.edu
Sun, 6 Feb 2000 08:18:41 -0700 (MST)

These are problems with monolithic virtual machine architectures. I would
say the next best thing in virtual machines will be a virtual machine that
runs on top of the network, not the device. When the device plugs into the
network it identifies itself to the network and is able to be used by the
network's virtual machine. By identifying the device, the network can use
the appropriate device drivers (including the appropriate compilation
driver), sort of like how plug and play works (also this sounds a bit like
Jini which is only half baked on these issues). Disconnected devices (like
a PDA) make the problem a bit more complicated and interesting.

I agree our current notions of operating systems and virtual machines are
unmanageable, even for large computing devices like PCs. This is only
going to get worse as the devices become smaller, more numerous and more
special purpose. These problems are part of my PhD research focus. 


Sean McDirmid

On Sun, 6 Feb 2000, Stephen Pelc wrote:

> The problem with using external resources is that the node 
> vendor now has to expose implementation details. In many cases a 
> back-end linker is now required, and the program delivery 
> mechanism now has to know about the target node. This defeats 
> the primary objective of many virtual machine implementations 
> for cost-sensitive applications. Having program servers is fine 
> in a workstation/PC environment such as ANDF and its incarnation 
> in OSF, but becomes a maintenance and certification nightmare 
> for embedded systems.
> I believe that tuning the VM for easy back-end compilation is a 
> more commercially viable approach.
> --
> Stephen Pelc, sfp@mpeltd.demon.co.uk
> MicroProcessor Engineering Ltd - More Real, Less Time
> 133 Hill Lane, Southampton SO15 5AF, England
> tel: +44 1703 631441, fax: +44 1703 339691
> web: http://www.mpeltd.demon.co.uk