Special Features:

THE GRID: DEFINING THE FUTURE OF THE INTERNET (Part 2 of 3)
By Wolfgang Gentzsch, MCNC Grid Computing & Networking Services

The Three Waves Of Grid Computing

Grid computing evolved from high-performance and distributed computing in the 1990s. The evolution was driven primarily by the ever-growing need for computing resources; the availability of increasingly more powerful technology for networking, servers, middleware and applications; and the development and widespread acceptance of the Internet and the World Wide Web. Today, as previously with the Internet and the Web, Grid computing can be viewed as evolving in three waves: the Research Wave, the Business Wave and the Consumer Wave.

The Research Wave started in the mid-'90s with projects such as I-WAY and the Information Power Grid in the United States, Uniform Access to Computing Resources (UNICORE) in Europe, and NINF in Asia, and is now in its full development. Researchers in universities and in industries are currently developing Grid standards and Grid middleware, and they are building Grid testbeds for running complex, Grid-enabled applications.

While we are still working on user-friendly Grid "plug-ins" for common applications, Grid middleware, such as the Globus Toolkit and similar distributed resource frameworks, are certainly in their second generations and are providing increasingly more interoperability. This achievement is primarily the result of a hard-working, collaborating Grid community united by the Global Grid Forum (GGF), a community-initiated group of thousands of individuals from research and industry leading the global standardization effort for Grid computing via the creation and documentation of "best practices" -- technical specifications, user experiences and implementation guidelines.

Enterprise users and technology vendors are now becoming aware of Grid computing and its benefits. Awareness and interest are being driven by the growing number of successful Grid projects in the global research community. In these research projects, benefits including remote access, improved resource utilization, collaboration in virtual organizations and increased productivity have been clearly proven. Today, these benefits are demonstrated in thousands of Grid-like production environments and intelligent clusters of resources distributed within enterprises all across the world -- enabled by distributed resource management software including Load Sharing Facility (LSF), Portable Batch System (PBS), Grid Engine and, on a more global level, by Globus, Avaki and the like.

In addition, vendors including HP, IBM, Oracle and Sun have started Grid projects which primarily focus on reducing cost and complexity in the enterprise data center, with initiatives such as the Adaptive Enterprise, Autonomous Computing (on demand) and N1. Recently, IT vendors have taken another serious step towards developing enterprise Grid solutions by establishing the Enterprise Grid Alliance (EGA), a consortium of vendors and customers, with a pragmatic focus on the adoption and deployment of Grid standards and technologies, including interoperability, for enterprise solutions.

The Business Wave, or the second wave of Grid computing, has begun. The aforementioned developments aimed at providing Grid solutions for businesses and involvement of mainstream technology vendors are strong indicators and evidence of this. In the next few years, we will certainly see a wide variety of efforts in industry to implement Grid standards and interoperable technologies which will allow any company to conduct business, over a worldwide and often specialized and customized Grid, in a user-friendly, reliable, efficient and secure way.

The Consumer Wave, or the third wave of Grid computing, is on the horizon. Grid computing will deliver unprecedented benefits to consumers and may equal in importance the Internet today. An important prerequisite to enable development of the Consumer Wave is a strong and healthy global Grid community, driven by the joint interest, dedication and working commitments to build the next-generation IT infrastructure for research, government, industry and consumers.

Grid Standards

By its very nature, a Grid is heterogeneous, often multi-domain, and responsible for connecting resources over a wide area, including worldwide. The network, compute, middleware, data and application components are from many different vendors. To be able to build these Grids easily, and change them on the fly to adjust to ever-changing needs and requirements, Grid components need to possess a high degree of interoperability and facilitate robust communication.

For this to happen, we need standards for: designing a complete Grid architecture; data access and integration; Grid security; resources discovery and monitoring; scheduling and resource management; applications and programming models; and much more. This is the task of the GGF. Other organizations which are important to GGF are: IETF for Internet standards, DMTF for distributed management standards, OASIS for e-business standards, WS-I for Web services interoperability, and W3C for interoperability for the Web.

To complement GGF's efforts, especially in the areas of implementation of Grid standards and interoperability of products, the Enterprise Grid Alliance (EGA) has been founded by IT vendors such as EMC, Fujitsu, HP, Intel, Network Appliance, Oracle and Sun.

This conglomerate of vendors, customers, standards bodie, and service organizations, with their somewhat overlapping agendas, demonstrates the grand challenge to develop lasting standards for the next generation of our Advanced Web and The Grid.

Grids Today

Though we still might be five years away from the "utility" Grid model, the evolution has begun. Grid computing is a reality today, and the list of early applications driving the development of Grid technology is substantial. It started with large-scale simulations in research, engineering, biology and drug design. Now the benefits are being seen in the sharing of digital content, application service provisioning, data mining for research and business applications, collaborative design, remote use of experimental instruments, remote medical diagnostics and synthesis of large distributed data sets. The list goes on and on.

According to Duncan Ross, director of IT at Cognigen Corp, cluster Grid software has enabled him to achieve a 10-to-1 reduction in hardware costs while doubling usable CPU power. Since implementing the Grid, Cognigen scientists have found that they now have one extra hour every day to work on other initiatives.

But while we are building Grids today, there are still only a few standards that are commonly agreed upon. Consequently, today, much of the work required to set up and run a large prototype Grid is manual.

If the main requirements are high utilization, management and easy access to the resources in a single department, then obviously, complexity is low, and a distributed resource management system will address the needs.

In an enterprise Grid that includes resources from different departments or servers consolidated into one powerful cluster for all departments or business units, complexity is much higher. This requires "supervisor" software to take care of fair sharing of all resources, according to the organization's policies.

At the next level, an enterprise can leverage its compute and data resources which may be distributed on a global basis. This requires the next layer of Grid software. Grid software in this category provides the capability to handle distributed data sets and to create and manage a single global name space, security, authentication and authorization.

Grid computing already has been implemented in numerous industrial settings, on both a departmental and an enterprise level. Grids are also coming into the commercial setting through third parties. Many ASPs and ISPs are implementing Grid-enabled access to applications.

Unlike technical computing, however, where the key demand is mostly raw system performance and better resource utilization, the demands in commercial environments also cover scalability, manageability, availability, reliability and transparent access to all resources of an enterprise.