Computer Systems Project Proposal
Sean Colyer
Period 1
9/29/05

I.  Creation of software and grid in a multiple operating systems distributed 
computing environment.

II. Problem Statement. 
There are many computers that use nowhere near their full potential, they could
be networked to create a powerful grid and take use of many computers fading out
of date.

III.    Purpose. 
This project has a number of reasons and goals. I would be working to develop 
and study network and processor management that would work in a number of 
environments. I will explore current grid computing techniques and explore 
ways to tie them together so that they would work in different operating 
systems. This project could help small networks with older computers work 
towards using machines as one common machine without seriously changing their 
operating systems and style of network.  A grid computer in the Computer 
Systems Lab would allow for the creation of a mild supercomputer for students 
to study and use for higher level computer science and math projects. If 
created effectively, a grid computing network of this nature could run on an 
environment with a wide range of computers without forcing large changes and 
done through only running a small program that while not in use would use 
minimal cpu power.

IV. Scope of Study. 
To create an inter operating system compatibile grid application, the only 
practical solution is to write it using Java. Using Java, I will be able to 
tweak the program for the different operating systems while keeping it 
contained within the same project files. Ideally, I would write the program so 
that I do minimal algorithm writing and instead write end use applications that
will allow computers to plug into a variety of grid networks (Apple's Xgrid, 
and OpenMOSIX for Linux).

V.  Background and review of current literature/research in this area. 
Distributed and grid computing is currently a very popular topic. Virginia 
Tech's cluster of 1100 dual processor Mac G5's went to show how effective grid 
computing could be.  Apple releases Xgrid with its new OS X, 10.4. Xgrid can 
run on any mac running 10.4.x and allow it to easily bring nodes in and out of 
a grid computing, creating a virtual on the fly grid computer, the problem is 
that it is limited to Mac computers. Apple's recent announcement of their 
transition to the x86 architecture it is viable that Xgrid could be ported to 
other operating systems. OpenMOSIX is a popular kernel extension for Linux that
creates excellent load balancing and easily creates a grid computer for those 
who are willing to use the rather limited conditions it allows. There has been 
virtually no creation of a grid computing network for Windows and for use in 
multiple operating systems. An effective product that could be used in a 
variety of environments could transform workplaces with large numbers of 
desktop using a fraction of their potential.

VI.   Procedure and Methodology.
The first part of my project is to work on creating an Xgrid server in Java, 
there is a project already that has created an Xgrid client in Java which can 
be used in a range of operating systems. I'm going to try to create a program 
that uses the network interface from Xgrid and then use slightly modified load 
balancing from other grid computing projects. The more computers that I can get
for a testing environment, the better, any range of ex-workstations could be 
incorporated into the grid. My testing will be based primarily starting with 
2 nodes and a central node, I will try examining the way OpenMOSIX works on 
this cluster and then try creating and implementing software for Windows.
                            
VII.    Expected Results & Value to Others
This project could help create a viable solution to turn many desktops into 
nodes in a cluster to create grid computers with lots of processor power 
without drastically changing the work environment. Eventually I hope to be able
to showcase the project by using a variety of computers in a cluster, using 
grapsh and charts of processor use and task management. I think the project to 
be done effectively will probably take all year, but is a good project because 
I can spend more time, the research will not really be complete but can create 
a product. I will track progress in a daily log.