More than 400 scientists gathered at the Global Grid Forum this week to discuss what may be the Internet’s next evolutionary step.
Though distributed computing evokes associations with populist initiatives like SETI@home, where individuals donate their spare computing power to worthy projects, the Grid will link PCs to each other and the scientific community like never before.
IBM’s Brian Carpenter suggested “computing will become a utility just like any other utility.”
Carpenter said, “The Grid will open up … storage and transaction power in the same way that the Web opened up content.” And just as the Internet connects various public and private networks, Cisco Systems’ Bob Aiken said, “you’re going to have multiple grids, multiple sets of middleware that people are going to choose from to satisfy their applications.”
As conference moderator Walter Hoogland suggested, “The World Wide Web gave us a taste, but the Grid gives a vision of an ICT (Information and Communication Technology)-enabled world.”
Though the task of standardizing everything from system templates to the definitions of various resources is a mammoth one, the GGF can look to the early days of the Web for guidance. The Grid that organizers are building is a new kind of Internet, only this time with the creators having a better knowledge of where the bottlenecks and teething problems will be.
The general consensus at the event was that although technical issues abound, the thorniest issues will involve social and political dimensions, for example how to facilitate sharing between strangers where there is no history of trust.
Amsterdam seemed a logical choice for the first Global Grid Forum because not only is it the world’s most densely cabled city, it was also home to the Internet Engineering Task Force’s first international gathering in 1993. The IETF has served as a model for many of the GGF’s activities: protocols, policy issues, and exchanging experiences.
The Grid Forum, a U.S.-based organization combined with eGrid – the European Grid Forum, and Asian counterparts to create the Global Grid Forum (GGF) in November, 2000.
The Global Grid Forum organizers said grid communities in the United States and Europe will now run in synch.
The Grid evolved from the early desire to connect supercomputers into “metacomputers” that could be remotely controlled. The word “grid” was borrowed from the electricity grid, to imply that any compatible device could be plugged in anywhere on the Grid and be guaranteed a certain level of resources, regardless of where those resources might come from.
Scientific communities at the conference discussed what the compatibility standards should be, and how extensive the protocols need to be.
As the number of connected devices runs from the thousands into the millions, the policy issues become exponentially more complex. So far, only draft consensus has been reached on most topics, but participants say these are the early days.
As with the Web, the initial impetus for a grid came from the scientific community, specifically high-energy physics, which needed extra resources to manage and analyze the huge amounts of data being collected.
The most nettlesome issues for industry are security and accounting. But unlike the Web, which had security measures tacked on as an afterthought, the Grid is being designed from the ground up as a secure system.
Conference participants debated what types of services (known in distributed computing circles as resource units) provided through the Grid will be charged for. And how will the administrative authority be centralized?
Corporations have been slow to cotton to this new technology’s potential, but the suits are in evidence at this year’s Grid event. As GGF chairman Charlie Catlett noted, “This is the first time I’ve seen this many ties at a Grid forum.”
In addition to IBM, firms such as Boeing, Philips and Unilever are already taking baby steps toward the Grid.
Though commercial needs tend to be more transaction-focused than those of scientific pursuits, most of the technical requirements are common. Furthermore, both science and industry participants say they require a level of reliability that’s not offered by current peer-to-peer initiatives: Downloading from Napster, for example, can take seconds or minutes, or might not work at all.
Garnering commercial interest is critical to the Grid’s future. Cisco’s Aiken explained that “if grids are really going to take off and become the major impetus for the next level of evolution in the Internet, we have to have something that allows (them) to easily transfer to industry.”
Other potential Grid components include creating a virtual observatory, and doctors performing simulations of blood flows. While some of these applications have existed for years, the Grid will make them routine rather than exceptional.
The California Institute of Technology’s Paul Messina said that by sharing computing resources, “you get more science from the same investment.”
Ian Foster of the University of Chicago said that Web precursor Arpanet was initially intended to be a distributed computing network that would share CPU-intensive tasks but instead wound up giving birth to e-mail and FTP.
The Grid may give birth to a global file-swapping network or a members-only citadel for moneyed institutions. But just as no one ten years ago would have conceived of Napster — not to mention AmIHotOrNot.com — the future of the Grid is unknown.
An associated DataGrid conference continues until Friday, focusing on a project in which resources from Pan-European research institutions will analyze data generated by a new particle collider being built at Swiss particle-physics lab CERN.