What Is the Grid?
A Grid computer is a hardware and software system that integrates a collection of distributed system components (e.g., computer systems, storage, etc.) making them appear to the user as a single, large "virtualized" computing system. The basic "single system" concept may be applied to the construction of centralized "cluster" computers (multiprocessor systems consisting of colocated computers and storage) or a networked grid of geographically dispersed computers, instrumentation, or other resources.
With Grid computing, computer systems and other resources aren't necessarily constrained to be dedicated to individual users or applications, but can be made available for dynamic pooling or sharing to meet the changing needs of the organization. By using the Internet, grid-based resource sharing and collaborative problem solving can be extended to multi-institutional "virtual organizations."
The name "Grid computing" was chosen because the concept is analogous to the electrical power grid, which satisfies customer demand for power by transparently drawing on a distributed grid of power-generating resources shared by all users.
Benefits of the Grid
Because it is evolving as a very flexible, perhaps even all-encompassing, distributed system technology, Grid computing offers a number of potential uses and benefits that can be broadly categorized in the following way:
High Performance Computing (HPC): For applications that lend themselves to parallel computing, the Grid offers the possibility of executing computationallyintensive applications on networked computer arrays consisting of numerous commodity or specialized systems. Computational Grids can offer significant advantages in both price/performance and in maximum performance over more conventional types of supercomputers. As a result, the Grid makes HPC accessible to more enterprises, accelerates the availability of results of computationally intense analysis needed for product research and development, and allows scientists to solve "grand challenge" problems that were too large for conventional supercomputers.
In its current stage of evolution, most applications of the Grid fall into the HPC classification. This is due to the fact that Grid computing arose out of the need for more cost-effective HPC solutions to address critical problems in science and engineering. The initial adoption of the Grid by commercial enterprises has continued to focus on HPC because of the high return on investment and competitive advantage realized by solving computeintensive problems that were previously insolvable in a reasonable period of time or cost. The HPC Grid has successfully addressed a wide range of computational problems in the following fields:
- Climate/weather/ocean modeling and simulation
- Computational chemistry and materials science
- Environmental quality modeling and simulation
- Military forces modeling and simulation
- Internet search engines
- Pharmaceutical research
- Simulation and verification of electronic and mechanical design
- Seismic processing and interpretation
- Signal/image processing
- Modeling of financial portfolios and markets
Data Federation and Collaboration: The Grid also allows a federated approach to data integration where data from different sources (relational databases, files, or application data) can be consolidated in a single data service that hides the complexities of data location, local ownership, and infrastructure from the consuming application. With data federation, the data remains in place at its source, with no disruption of local users, applications, or data management policies. Integration of data from multiple sources and locations facilitates a wide range of integrated applications, including corporate performance dashboards, marketing analysis tools, customer service applications, and data mining applications. Because the data resource is accessed as a service on the network, minimal modifications of existing data publishing or data-consuming applications is required.
Resource Allocation and Optimization: Most desktop systems and servers are idle more than 80% of the time because computers have traditionally been rigidly dedicated to certain sets of users and applications. Grid computing provides a virtualization framework that allows flexible sharing of computing and storage to improve resource utilization. In the simplest example, a batch job can be transparently allocated to an idle server in the pool of resources. Alternatively, both the application and the job could be transferred to an idle server. In addition, if virtual machine software is installed, a single physical server can even be configured on the fly to run different operating systems and applications. In a similar way, idle storage capacity of Network Attached Storage, or SANs, could be utilized for data storage.
