Grid Networks: The New Computing Infrastructure

Abstract

Grid networks are an evolutionary descendant of both parallel and cluster computing architectures. They have much in common with the electric grid in that processing power is designed to be distributed ubiquitously within computer networks, without information as to where the source is. Also like the electric grid, grid networks are designed to connected heterogeneous systems in a reliable and dynamically changing way. Grid networks have required the development of new software tools for resource allocation and management, resource information publishing, security, and communication just to name the primary functions. These services have all been incorporated into a functional open-architecture, open-source project called Globus. Both Internet infrastructure development and grid network infrastructure development began with government research initiatives which eventually led and are leading to government-induced privatization of both networks respectively. Firstly, standard technical procedures and protocols were set network wide. Secondly, legislation stimulated private sector growth to induce commercialization of the infrastructure. However the two projects differ in that grid infrastructure development has been able to use the existing Internet hardware infrastructure base, commercial sector, and established nongovernmental collaborative research organizations as springboards. We are currently in the midst of a privatization phase of grid networks. An international organization called the Global Grid Forum, consisting of world grid network researchers, is likely to soon standardize Globus or some variation of it as the globally accepted wide-area distributed supercomputing platform. An American government initiative to lease the processing time of a Globus-operated supercomputing network called the National Technology Grid, is likely to be followed by legislation to fully privatize the network. However, due to existing large commercial roles in the project, this legislation may only be a formality. This new commercial information processing infrastructure will be the source of inexpensive supercomputing power provided and used largely by scientific and medical research communities, accelerating the advancement of all research that requires the use of high-performance computers. Differing security and performance requirements may cause the eventual partitioning of the grid into networks organized according to processing requirements rather than national or political interests (e.g. industry, academic, commercial, financial, entertainment).