Distributed Computing Overview
A short summary of distributed computing.
Distributed computing is an all encompassing term for concurrent (i.e. simultaneous) computation across autonomous computers. While this definition could be applied to many forms of computation, including cluster computing and super computing, generally distributed computing refers to what is more specifically grid computing.
In grid computing, a number of independent machines are connected via a network and act together to perform very large tasks. Unlike cluster and super computing, however, each computer within the grid is independent and cannot directly access the resources (CPU, memory, peripherals, etc.) of other computers on the grid. Grids are managed by middle-ware, such as Xgrid or Globus, that act as glue between the computers on the grid.
There also exist types of middle-ware that do not create grids in the traditional sense, but instead schedule batches of jobs to computers that would otherwise stand idle while their user wasn't using its resources. The most prominent example, and one that is supported for the arts, science and engineering community by ASUnix, is Condor. An interesting feature of Condor is its ability to run as a 'Personal Condor' which only uses the user's workstation for computation but still allows the user to harness the power of Condor's batch processing system for running automated simulations. Another supported system for capturing unused computing resources on PCs and workstations is BOINC, which is the middle-ware layer for popular distributed research platforms SETI@home and Milkway@home.
If researchers or departments would like to utilize the resources available to them under Condor, Personal Condor or BOINC, they should contact ASUnix in order to initiate the process of developing a distributed computing platform to suit their needs.