Special Features:

THE GRID: DEFINING THE FUTURE OF THE INTERNET (Part 1 0f 3)
By Wolfgang Gentzsch, MCNC Grid Computing & Networking Services

Grid computing is no longer the exclusive realm of researchers seeking to harness enough compute power for massive computational challenges. Commercial enterprises are now focused on the tremendous benefits that Grid computing will yield. Yet, it's still in an early stage of its evolution. "The Grid" has the potential to become a global Web of ubiquitous electronic services which will improve infrastructure utilization, increase data access/integration, enable new levels of communication and provide for the creation of new applications.

So, what is Grid computing, and what is "the Grid"?

In 1998, Grid gurus Ian Foster and Carl Kesselman defined the Grid as a hardware and software infrastructure that provides dependable, consistent and pervasive access to high-end computational resources. Their definition in 2000 complements this, stating that Grid computing is concerned with coordinated resource sharing and problem solving in dynamic, multi-institutional virtual organizations. When we mention resources these days, we include the network, distributed computers which communicate among each others, distributed storage systems, mobile devices, including those on our bodies to monitor our health status, scientific experiments, instruments, sensors and sensor nets, databases, software applications and more. A great example of scientific experimental facilities shared over a Grid is the NEES Network for Earthquake Engineering Simulations project, where scientists all over the United States have access to shake tables, tsunami wave basins, geotechnical centrifuges, field testing and large-scale structural testing equipment.

SearchCIO.com states that Grid computing is applying the resources of many computers in a network to a single problem at the same time -- usually a scientific or technical problem that requires a great number of computer processing cycles or access to large amounts of data. This definition, however, seems limited in its reference to a single problem. In reality, at least in today's Grid environment, we see job mixes of multiple problems.

For Oracle's Larry Ellison, the Grid is "capacity on demand." Plug another server into the Grid and the application runs faster and more reliably." For Sun CEO Scott McNealy and for John Patrick, IBM's vice president for Internet strategies, Grid computing is simply "the next big thing," and I can agree with that one, too.

The definition of Grid computing keeps shifting, and that's not a bad thing at this stage of the game. The technologies and standards surrounding Grid computing are still evolving, as they should. One thing that hasn't changed -- and isn't likely to, for semantic reasons -- is the underlying analogy to the electric power Grid, with ubiquitous service. It's embedded in the name "Grid computing."

What if computing power were a utility, like electricity, telephony or running water? That remains the concept that appeals to a wide range of interested parties, including researchers, technologists, investors, business professionals and, most of all, laymen. The utility model is what captured our imaginations in the first place.

So what are the salient characteristics of a utility? Three of the most prominent are:

That's why I like using a simple, "utilitarian" definition of Grid computing. Look no further than the recent and rapid evolution in the wireless telephone industry to see how difficult it is to apply rigid categories in an area of technological innovation.

In the early stages of the evolution of wireless telephones, nobody expected that the technology would extend to Web browsers and cameras, or that mobile telephones would morph into a combination of telephone and personal digital assistant (PDA) functions. It is this kind of innovation and subsequent evolution in the application of Grid computing that we will experience.

Today's challenges and main applications of Grid technology barely hint at its full potential.

Grid vs. Internet

How does the Grid compare to the Internet? The Internet primarily relies on people (producers) who enable the provisioning, accessing, sharing, exchanging and managing of information to end users via hyperlinks, Web sites, plain text and more. The Internet itself does not automatically handle these functions or this information. Thus, the Internet primarily is a network with static compute and data resources in which information flows from the producer to the consumer (and sometimes back). Because of its static nature, applications (mostly information) cannot easily adapt to changing network or server conditions; neither can resources adapt to changing application requirements.

In contrast, far beyond the capabilities of today's Internet, the Grid enables provisioning, accessing and sharing, as well as adding, removing and managing of resources which are dynamic components of the Grid and can be managed fully and automatically through policies, application requirements or end-user interaction.

Thus, a Grid can be programmed to perform certain tasks -- for example, computer simulations, conducting experiments or database searches -- to automatically include more resources when needed or to broker resources according to users' requirements, such as time deadline, budget, priority, level of security and other capabilities.

As a community of resources, any Grid can be compared with a community of people addressing a specific need or task. They are distributed, heterogeneous, coordinated and managed. Each person, or resource, is intelligent, specialized, able to communicate and, above all, can perform a task as part of a larger solution (e.g., workflow).

This "human Grid" usually is extremely self-adaptable. In situations in which someone suddenly doesn't perform as expected, such as being absent due to illness, often the first noticeable effect is that the whole project slows down. Then, someone else fills the gap to bring the project back on track. The community adapts and provides self-healing and the necessary support to achieve the goal. The entire process realigns itself along the ever-changing infrastructure.

Ideally, a Grid of computing resources behaves exactly the same way. To perform a specific task, such as a complex simulation in the automotive industry, many different computing resources in a Grid work together in a coordinated and managed way. As in a "human Grid," the computing Grid can dynamically change, for a multitude of reasons. To guarantee continuous execution and accomplishment of the task, the Grid has to adapt itself automatically to any change without affecting the (execution of the) application.

Today, if an Internet information server crashes, we will not be able to access information that resides on the server, because computing, storage and network resources are statically assigned to the application. The only option is to wait until the server is up again.

In a Grid, if a server goes down, this will immediately be recognized by the Grid master node, by an intelligent peer node or by the application itself, and the peer or another healthy node will be brought into the Grid to take over. From the perspective of the end user, the entire process in the Grid is seamless and transparent.

The Grid resource infrastructure provides self-adaptation capabilities. However, as with "human Grids," we also have to build intelligence into the projects themselves to optimally adapt to our changing human infrastructure. The wide field of mechatronics, which combines mechanical with electronic (intelligent) components, is full of great examples. Therefore, applications themselves should also be intelligent, aware and self-adaptable to be optimally efficient in a Grid environment.

The Net Effect

Within the past decade, we've witnessed a convergence of exponentials -- Moore's Law meets Gilder's Law meets Metcalfe's Law -- in which processor speed, bandwidth and the value of networks have all increased exponentially in the past. Some call it the Net Effect: advances amplifying advances.

Those advances are taking us toward the Grid -- the next logical step from the Internet and the World Wide Web, to the infrastructure that connects distributed computers, storage systems, mobile devices, instruments, sensors and sensor nets, databases and software applications, while providing uniform access for the local, regional and global user community.

The Grid is the product of evolution, dovetailing the next generation of the Internet and the World Wide Web, an advanced form of today's Web. The Grid will be used for computing, collaboration and communication. And like the current Web, the Grid will be ubiquitous. It will simply become the basic IT infrastructure for many businesses and applications, within research groups and commercial enterprises, and, eventually, for consumers.

The Grid will be available to anyone, anywhere, anytime, for any kind of personal or professional purpose. There are, however, still years to go until we simply switch on our appliance and use any information, Web service, computing service or whatever service -- all available over the Internet in ways that are just as pervasive as today's Internet.

About Wolfgang Gentzsch

Wolfgang Gentzsch is the managing director at MCNC Grid Computing & Networking Services, an independent, nonprofit, advanced technology research and service center that develops, evaluates and deploys Grid computing and advanced networking solutions. Wolfgang joined MCNC in April 2004. He was formerly senior director of Grid computing at Sun Microsystems Inc. In July 2000, Sun acquired his company, Gridware, a spin-off of Genias Software, which he founded in 1990. From 1985 to 2000, Wolfgang was a professor of Mathematics and Computer Science at the University of Applied Sciences in Regensburg, Germany.