Grid Computing

Definition

Grid computing is a distributed computing model that pools the processing power, storage, and other resources of many independently owned and geographically dispersed computers to work together on large computational tasks.

Overview

Grid computing emerged in the 1990s and early 2000s from academic and scientific computing communities that needed more processing power than any single institution could afford. Rather than building one enormous supercomputer, a computing grid links together many separate computers — often owned by different organizations — over a network, coordinating them to divide up and execute pieces of a large job. Well-known early examples include SETI@home and Berkeley Open Infrastructure for Network Computing (BOINC) projects, which harnessed idle cycles on volunteers' home computers for scientific research. A grid differs from a single High-Performance Computing (HPC) cluster in that its nodes are typically heterogeneous, loosely coupled, and administratively independent, connected over standard networks rather than a specialized low-latency interconnect. Middleware software handles job scheduling, resource discovery, authentication across organizational boundaries, and reassembling results as individual nodes complete their portion of work. Because nodes can join, leave, or fail unpredictably, grid systems are built with significant tolerance for partial failure, unlike tightly synchronized Parallel Computing systems. While cloud computing has largely absorbed the commercial use cases grid computing once served — renting elastic, centrally-managed compute instead of federating volunteer or institutional resources — the underlying ideas of grid computing (resource federation, distributed job scheduling, and large-scale collaborative computation) still inform modern Distributed Systems design and remain in active use in scientific research, such as the Worldwide LHC Computing Grid used to process particle physics data.

Key Concepts

• Federates computing resources across multiple independent administrative domains
• Nodes are typically heterogeneous in hardware and geographically distributed
• Middleware handles scheduling, authentication, and resource discovery across the grid
• High tolerance for individual node failure or disconnection
• Jobs are divided into independent work units distributed to available nodes
• Often built on volunteer or shared institutional computing capacity
• Loosely coupled compared to tightly synchronized cluster computing

Use Cases

Frequently Asked Questions

Related in Computer Science