Grid Computing Overview
Grid computing is a new model for managing large scale computational and data resources. The vision of the designers of grid computing is that computing resources should be like electricity - plug it in, turn it on, and it works. While we have not yet achieved that vision we have been able to focus large amounts of resources on particular problems. Typically grids use the resources of separate but connected computers. The computers may be your basic desktop machine, larger servers or super computers.
You can find a good overview of grid computing at Wikipedia:Grid_Computing
Grids come in different forms, which may be combined. A computational grid is designed to deliver larger computational resources than would otherwise be available. The design of the grid tries to hide as much detail as possible from the user making it as easy to use hundreds of machines as to use one.
A data grid manages large amounts of data. It often makes the data appear as if it is a single resource, even though it may be stored at many locations. Data grids are often enhanced with additional services such as the ability to search and retrieve selected data.
An Access Grid
is designed to facilitate collaboration between distributed groups of people. The Access Grid supports audio and video feeds between sites as well as the display of presentations or executing applications. The three grids may be brought together when a distributed group of researchers watch the output of an application executing on a computational grid. That application may be drawing data from a data grid.
Grids and Certificates
Grids are designed to share resources located at different institutions. For example a user, Sally, at VUW might wish to perform a computation on Canterbury's BlueFern using data retrieved from the GeoNet database at GNS. This will require the data to flow over KAREN between GNS and Canterbury and the results to be returned to VUW. Sally will need to be authenticated to both use Canterbury's BlueFern and to retrieve data from GeoNet. To do this Sally will need to obtain a certificate from the appropriate certificate authority. You can read about Sally's example
to see how certificates are used.
Can the Grid Help You?
If you do a lot of computation it will be worthwhile investigating the grid as a new tool. A key benefit users mention is the ability to obtain results in hours that might take days or weeks to process on a single computer.
Victoria's cycyle-stealing grids are made up of networked workstations. The processors are independent and do not share any memory. Applications that are most suitable for these grids can be broken down into components that work on separate data sets. An obvious example is an application where a large set of similar yet independent runs are required.
A limited amount of communication between processes, such as communicating boundary conditions, can be tolerated without degrading performance.
Many applications can be moved to a grid environment with minimal work. Other applications that deal with a large monolithic dataset, may require rewriting to use domain specific libraries, or a change of algorithm or approach to take advantage of the parallelism of the grid.
VUW staff and students have used the grid for a variety of e-research projects