Leonardo Digital Reviews
 LDR Home  Index/Search  Leonardo On-Line  About Leonardo  Whats New






 

LDR Home

Current Reviews

Review Articles

Book Reviews Archive: July 2000 to October 2002

Book Reviews Archive: 1994 to May 2000

The Grid: Blueprint For A New Computing Infrastructure

edited by Ian Foster & Carl Kesselman

677 & xxv pages including index, bibliography, & glossary.

Reviewed by Curtis E.A. Karnow


Apres Moi, Le Deluge

Over thirty contributors make this a comprehensive, massive tome on the projected next leap of the global computing infrastructure. Alluding to the power grid, the authors make the case for a computational grid, able to harness massive power and databases fed into the smallest local terminal, much as the power of nuclear reactors and water dams light the bulbs on our scattered desks.


The collection has four stated goals--a manifesto urging the construction of the grid; a design blueprint, a user guide, and a research agenda. It does all these, admirably. It will likely serve best as a text for courses in networking and distributed computing and reference work. But more importantly, this books is a beacon to those of us who follow the development of computing, who prognosticate, and wonder what the Internet will be, and hope how to leverage the stunning power of networked computers to qualitatively new uses.


But among the lessons here is how very hard it is to make a network, a true network of particles of every shape and size. Because we will not tear down the present infrastructure for a new one, inconsistent systems must be harnessed. And these are inconsistent in so many ways- different architecture, languages, operating systems, security systems, memory usage, data structures, and on and on. The challenge is to flow smoothly data and operations from massively parallel supercomputing systems all the way down to handheld machines; all the while maintaining data integrity and security, speed of access and seamless environments, fault tolerance, solving performance bottlenecks, and so on. The authors here do not shy away from the issues: the chapters here outline potential solutions, citing existing architectures, applications that are up and running, operational high bandwidth, very high speed wide area networks, and the development of the Globus Toolkit which provides basic grid services.


The book describes grid-enabled applications such as weather forecasting, database management with complex high dimensional data, modeling complex systems such as stars and ocean systems. These do not, however, address the promise of the Grid: such applications are already available to a significant degree at present supercomputer sites, and do not leverage distributed computing power for the individual. But one chapter, devoted to teleimersion, does signal that promise. To be sure scientific (including medical) visualization, architecture review and scenario simulation are furthered by teleimersion, but so is art, and human contact. The computing power, enormous throughput, and communications infrastructure contemplated by the Grid is essential to what we used fondly to call virtual reality--a reality that, if the authors of the Grid are right--will seamlessly augment this one.


It is no surprise that the new cannot be predicted. Technology does not only fulfil needs, it makes them; "we cannot image" the applications emerging from "reliability and location independence among personal digital assistants and networked computers" (534).


Of particular interest to those interested in distributed autonomous computing is one author's confirmation that an agent-based approach will likely be needed to manage complex highspeed widely distributed networks. In an operational Grid, "when we observe that something has gone wrong, it is generally too late to react." (93) Agents with adaptive behavior must be trusted to manage and fix the Grid. And I expect that human supervisory control will increasing high level. Predictable performance will become an increasing challenge; I hope the system treats us well.


This volume treads lightly or not at all on the social implications of the Grid. But it does provide the unvarnished detail of the Grid's slow eruption, protocol by protocol, tool by tool and program by program. Its detail conspires to make us believe it will happen.

 

top

 







Updated 1st June 2004


Contact LDR: ldr@leonardo.org

Contact Leonardo: isast@leonardo.info


copyright © 2004 ISAST