Grid Computing Basics

What is it?
Grid computing is, simply, the linking together of multiple computing resources – including processors and storage devices – to provide a much larger, more powerful single virtual computer. Grid computing offers a variety of benefits, including:

• Making individual applications run much faster.
• Allowing multiple applications to share computing resources.
• Increasing the use of all available computing resources.
• Removing knowledge of IT infrastructure from applications.
• Providing flexibility to add more resources to the grid as needed.

Who's using it now?
Grid computing has gained rapid acceptance in the commercial marketplace in a manner similar to the emergence of the Internet. Organizations with both large and small networks have been adopting grid techniques in order to reduce execution time and enable resource sharing. According to a recent Forrester Research report, the financial and insurance industries are leading the way, with public sector, manufacturing and telecom also showing growth in grid implementations and reporting interest in future use.

How does it work?
In general, grid computing uses parallel and distributed processing to dynamically share and aggregate multiple computer resources. Typically, arrays of computational power are constructed from a network of computers. The network of computers may consist of several computers in a single department, or it may consist of hundreds or even thousands of computers spanning multiple geographic locations. Applications are submitted to the grid, and a piece of software called "middleware" decides which specific machine in the grid should execute the application. This determination is made at run time and is based on the software and hardware requirements of the application as well as the availability of the computers in the grid.

How can you use it?
You can use a grid to get significant performance gains with individual applications by splitting these applications up into smaller pieces and running the pieces in parallel on a grid. Applications that are good candidates for acceleration via a grid implementation take many hours and possibly even days or weeks to run. In some cases, the job is so big that it cannot be completed at all, even given today's processor speeds. If a long-running application involves many runs of the same fundamental task, or breaks down into execution units or data subsets, then it may be a good candidate for acceleration via parallel processing on a grid.

In addition, grids can provide efficient utilization of a virtual pool of resources by multiple applications and multiple users. Concurrent users can submit multiple jobs to the grid and have them execute efficiently and effectively. This is due to the grid's ability to load balance, enable prioritization and enforce policies to provide better turnaround times for individual jobs, improved service levels and more efficient use of computing resources.

What role does SAS® play?
Because SAS is so analytically powerful, many SAS applications tend to be very data intensive, compute intensive or both. The performance of these SAS applications often can be improved dramatically by running in a grid environment. In addition, no business runs on a single application. A SAS grid environment provides a way to efficiently manage multiple SAS applications being run by multiple users and possibly many business units within an organization.

Implementing a SAS grid solution has many advantages:

• Several SAS solutions are grid-automated, which means they automatically generate SAS programs that can be executed in parallel on a grid.
• SAS' grid distribution logic can be embedded directly into SAS applications to execute subtasks in parallel on the grid.
• The grid logic can be used to easily submit multiple applications to the grid.
• Dynamic, resource-based load balancing is available through integration with Platform Computing, a leading grid middleware vendor.
• SAS' grid environment is administered from a single point, which reduces complexity and, therefore, cost.
• The grid solutions can leverage all of the platforms supported by SAS, including several flavors of UNIX, Windows and Linux. 
• SAS provides a visual interface for monitoring and managing SAS applications in a grid.

Bio: Merry Rabb is a Product Manager at SAS. Cheryl Doninger is a Research and Development Director at SAS.