GRID Computing SIG

From IlugCal

[edit] Introduction

According to wikipedia.org : Grid computing is an emerging computing model that treats all resources as a collection of manageable entities with common interfaces to such functionality as lifetime management, discoverable properties and accessibility via open protocols.
In common man's words, a computational grid lets users from different institutions to share resources. The Grid infrastructure allows sharing of geographically separated resources (for example clusters, tape drives, electron microscopes etc.). It is being succesfully used at a some very large projects and even in corporate sites to handle computing problems which are too big to be handled on a single machine (or even a locally built large cluster). What can be so big? Well, for starters, consider processing the data from the Large Hadron Collider at CERN ( yes!! just like the one described in Dan Brown's Angels and Demons). The sheer volume of data being produced is so large that it is impossible for any computer to process it. So CERN is using clusters all over the world (including clusters present at the Variable Energy Cyclotron Center, Calcutta), to crunch this data. How? The clusters are shared on Grids!! Seems interesting enough? Read on...

[edit] A Project Proposal: Deploying a demo Grid TestBed

Currently, complete Open Source Frameworks exist for implementation of fully functional production-strength grids. We can try out developing a few (at least two) small grids comprising of one clusters each and connect them to create a grid testbed.
This is a big job in itself, so we can approach it in small steps:

   1) Design the grid
       a) Design the network
       b) Decide on service distribution ( which grid services will run on which computers )
       c) Decide the Cluster gateway interactions ( whether cluster and Globus Gateway will lie on same host/ different hosts, 
what will be the network config between the cluster and the gateway etc)
   2) Build the clusters
   3) Install Globus on service providers
   4) Configure services
   5) Test run of the /bin/hostname command on the clusters through globus
   6) Use MPICH/G2 to compile BladeENC and run big mp3 encoding jobs on the Grids.

If all this is done we have a functional Grid.

[edit] Associated other projects

It will take over two hundred pages to briefly describe all the available tools for Grid computing. Most of these tools are distributed under open source licenses (and often under the GPL). So we can make improvements on them. Also we can develop different application layer software that can use the grid to process input data transparent to the user. In short there is great scope for development.

[edit] Associated Reading

1. One stop shop for Grid related information : Grid Computing Info Centre

2. Santanu and Kakoli have built a functional Grid testbed consisting of 2 small grids.
Their documentation can be found here.

3. Globus Toolkit is one of the most commonly used tools used to build computational grids. Read about it here.
This site also contains info about a lot of application level software for Grids.