Sep 28, 2003 (08:09 PM EDT)
Grid Computing: The Scoop On How It Works
Read the Original Article at InformationWeek
Hewitt Associates recently started using a grid of Intel-based servers to do pension-benefit calculations. Grid computing is relatively new; what is it and how does it work?
A grid is something like a roster in a team sport like football. It's a list of computing resources that a coach can send into the game. The players are servers, workstations, PCs, and, often, scientific instruments. Their plays are the software and networking equipment that connect them, and the field is the public Internet or a VPN.
The coach is a grid-management program--most commonly based on the Globus toolkit standards, though Hewitt uses a similar commercial software product from DataSynapse Inc.--which looks at all the computers on the roster and decides which can handle which tasks or whether two should team up on a task.
The football analogy isn't perfect. One big difference: There's no limit on the number of players that can be added to the grid. In a grid, if a quarterback can't throw the ball far enough, the coach can keep adding muscle until he can.
Grid-management software provides a common interface for sharing CPU cycles, files, and data over a wide area network, giving developers a standardized layer to write to without knowing the specs of each machine that might be brought online. All a user has to do is submit a calculation to a network of computers linked by grid-computing middleware. The middleware polls a directory of available machines to see which can handle the request fastest.
Paul Wanish, a grid architect in IBM's Design Center, puts the business uses for grid computing in three categories. Most common is the Big-Bang Grid: Blast out a massive amount of data, such as seismic data, and let the processors analyze it to get at the answer of where oil deposits may be. Second is the Monte Carlo Grid: Investment companies use them to do simulations--what-if scenarios to estimate probability in complex financial models. Third, and least common, are one-off calculation efforts such as Hewitt's pension-benefit calculations.
Grid computing and utility computing are sometimes confused. While a grid could be part of a utility computing infrastructure or an automated data center, a grid would be used to tackle specific workloads, typically scientific or mathematical. An IT utility provides all the basic IT resources (CPU, storage, networking, etc.) that a business needs to a run all its apps (manufacturing, ERP, human resources, and even Microsoft Office).