DAILY NEWS AND INFORMATION FOR THE GLOBAL GRID COMMUNITY / JUNE 9, 2003: VOL. 2 NO. 23    ( Table of Contents )

Special Features:

GRID COMPUTING: THE FUTURE IN REVIEW
By Wolfgang Gentzsch

At the FIRe (Future in Review) Conference in San Diego, May 19-22, panelists from Avaki, HP, and Sun were asked to detail recent milestones and progress made in the world of Grid Computing, as well as discuss the many challenges that lie ahead over the next five years. The conference -- initiated and organized by Mark Anderson, founder and publisher of Strategic News Service and President of Technology Alliance Partners (TAP) -- drew nearly 200 technology luminaries, such as Michael Dell and Ray Ozzie, to deliver their unparalleled views on the future of technology.

The special panel on Grid Computing was hosted by John Thompson, an industry veteran who has held senior technology and marketing positions at numerous companies, among them IBM, WhiteCross Systems, and PLATINUM Technology. John invited Wolfgang Gentzsch, Director of Grid Computing at Sun, Andrew Grimshaw, Founder and CTO of Avaki, and Greg Astfalk, Chief Scientist at HP to cover the day's topics, including:

• Defining Grids
• Grid Computing: hype or reality
• Compute Grids versus Data Grids
• Security
• Grid's major challenges

Excerpts from the Experts

Following are excerpts from the various view points shared by each expert panelists at the conference.

Grid Definition:

Andrew Grimshaw: A grid is the virtualization and secure sharing of resources (compute, data, storage, applications, etc.) both within an organization and between organizations.

Greg Astfalk: Grid renders virtually all of IT as a "Grid Service." This includes computer cycles, disk space, a file, a database, an application, an instrument, etc. Then the grid allows for operations on and with these grid services; define, register, discover, provision, share, remove, modify, etc. It will be the case that in time, the commercial space will be the more important one for Grid. Grid's heritage is scientific and it will continue to be viable there. However, the impact will be larger to the greater regime of "IT" and commercial computing.

Wolfgang Gentzsch: The Grid is a network infrastructure that enables remote, secure and transparent access to computers, applications, data, instruments, and more. I tend to differentiate grids into those built and used for research, and those for commercial use. Today, global grids, mostly for research, are in their prototype phase. However, already many so-called "mini- grids" can be found in the industry, in enterprises and in departments. Those Departmental and Enterprise Grids have many of the characteristics and benefits which we will find in future Global Grids, for example transparent access to resources, resource sharing, distributed resource management, and much better resource utilization. All of which results in better return on investment (ROI).

Grids: Hype or Reality?

Greg: Grid is over-hyped and this is a genuine problem today. Once end-users are disappointed by a new technology failing or under-delivering, recovery is very difficult or perhaps impossible. Grids need to be further deployed and tested and trusted before today's hype is justified.

Andrew: Grids are not all hype, grids are real today and both compute grids and data grids are being deployed commercially to solve real problems. It is not the case that grids have to be hard to use and deploy - and that our approach has been to keep them easy to use and as transparent as possible to end-users; for example, not require any code changes or re-compilation.

Wolfgang: There is definitely some hype around grids, especially when it comes to the final stage of "The Grid" as the next generation of the Internet and the World Wide Web. This "Great Global Grid" is still many years away. It will be the fundamental infrastructure to connect everybody and everything to the network. On the other hand, smaller grids, in departments and enterprises, with reduced capabilities, but still with enough benefits -- such as dramatically increased ROI and productivity -- are being built today. As an example, the Sun ONE Grid Engine distributed resource manager powers over 7,000 departmental and enterprise grids today.

Data Grids:

Wolfgang: Grids started out in the 1990s, originating from clusters of supercomputers, so-called metacomputers. Consequently, in the early days, the focus of grid research was more on the distributed computing side. But this changed a couple of years ago, with new applications -- like the European DataGrid for CERN's Large Hadron Collider experiment, or life sciences and financial applications -- that are powered by Terabytes and even Petabytes of distributed data.

Andrew: While compute grids have gotten most of the press - data sharing & management issues within and between organizations are often the most difficult challenge. One cannot simply buy a bigger cluster to solve the problem. One must address the fact that data is kept and managed in different locations (often basically data silos). By virtualizing access, Data Grids solve wide-area data management problems.

Security in Grids:

Greg: Security, privacy and trust are the top barriers to the wide deployment and adoption of Grid technology across communities, inter-enterprise and as a utility. Grid will allow people and servers you don't necessarily know to access an enterprise's systems and files. Trust is paramount before this will be allowed. However, it must be allowed for Grid to reach its potential.

Andrew: In particular, community security standards that are trusted by enterprises are a necessary condition for inter-enterprise grids (except in well-defined circumstances). Further, most of the grids that we have deployed have been enterprise grids - grids where all of the users and data sources are within the same organization - and usually within the corporate firewall.

Wolfgang: We will see companies offering different levels of security. Like Trusted Solaris, which is especially well-suited for classified government projects, we will need the highest level of security when we work with confidential information, especially over the Internet. While within an organization, or within an organization's department, security requirements might not be that stringent.

Other Challenges:

Greg: Grid is a complex software endeavor. It will take as much as a decade of hard work in development before we reach the notion of a utility Grid. Grid will evolve in a Darwinian evolution fashion as the challenges are overcome. Grid is not a silver bullet.

Andrew: And, there are the sociology issues. It is not just a technical problem to deploy a grid - it is a political and sociological exercise.

Wolfgang: Other challenges are grid standards and applications for the grid. Because grid standards are well taken care of by the Global Grid Forum, let's talk about applications for the grid. There is a common misconception that the only applications well-suited for a grid are those which can be decomposed (parallelized) over the many processors in the grid. That's only half true. Those applications build only a small subset of all possible grid applications.

A much larger class are so-called "parameter studies," in which hundreds or thousands of slightly varying parameters create hundreds or thousands of simulation jobs which you can throw at the Grid. Another even larger class are the job mixes coming from hundreds or thousands of users, which happens in universities as well as in industry.

But, over time, there will also be new applications, which will be especially suited for grids. Such applications are on-demand and real-time, adaptive and dynamic, complete workflows and workbenches, and collaborative computing frameworks. Finally, in the far future, we will be able to throw any application at The Grid.