It is now 10 years since the research community gave birth to Grid computing; 10 years since early projects like I-Way, Globus, Legion and Unicore gave us a hint to what we could do with this new technology; after the Grid “bible” from Ian (Foster) and Carl (Kesselman) started changing our minds and imagination; but the community is still struggling with its adoption, with the impact on our future and with its direction. Desperately, we are looking for unique signs from heaven, listening to the many Naradas, breaking our heads in helplessness.
And here comes one of these reliable recurrent kilometer stones (yes, I am back in Europe), giving us clear hints toward the future. I am talking about Grid 2005, this year's International Workshop on Grid Computing in Seattle, just before this year's Supercomputing. In its sixth year, this year's Grid workshop featured a program committee of over 100 computer scientists that reads like the “Who's Who” of the distributed computing community, selecting 30 papers and 18 posters out of almost 200 submissions. This is for sure one of the highest quality research conferences on Grid computing in the Universe, framed this year by visionaries (and veterans) like Dennis Gannon from Indiana University and his keynote on research challenges for the next generation of Grid systems, and Fran Berman (SDSC), Fabrizio Gagliardi (EGEE, now Microsoft), Carl Kesselman (UCLA/ISI) and Mark Linesch (GGF and HP) on the panel titled “What Will Grids Look Like in Five Years.”
So, what ARE the trends we are taking away from these heavyweights? Listening to the scientists, obviously, you recognize the current challenges and open questions. One area is coming up again and again — and it's data, data, data. I think I am still underestimating this single, largest challenge. How do we bring structure and order into this exploding data hurricane? Exponentially growing data, in many different formats, distributed over many sources and sites around the world, replicated for the sake of international collaborations in virtual organizations — all this requires sophisticated, scalable and robust data management capabilities. They have to take into account very heavy loads, performance bottlenecks, denial of service attacks and different policies, while still guaranteeing minimum service response time.
Another recurring topic is the wide field of security — protecting your core business, resources, data, applications, IP and other assets. Although authentication and authorization are well understood and implemented in simple environments, they are still a challenge for global or virtual organizations in research and industry. Policy-based distributed access control, security credentials and credential renewal for long-running jobs, identity management, group memberships, policy management, attribute-based mutual trust and trust building, resource negotiation, suspicion levels associated with requesters, and general scalability challenges result in central versus decentralized approaches. And, finally, firewalls and conflicting security policies play an important role and obstacle in today's collaboration among multiple organizations.
One important trend is the introduction of Web services as a means for standardized communications in services-oriented and interoperable systems, to expose application functionalities to end-users and to provide wrappers for legacy applications integration. Other contributions to the workshop concentrated on semantic Web technologies and Grids, complex fault scenarios in large distributed systems, (autonomous) failure detection and self-repairing systems.
The keynote speaker, Dennis Gannon, has been involved in so many distributed and Grid projects that to mention all would go far beyond the scope of this blog. So, HIS voice counts! Extrapolating from his lessons learned in Grid projects like I-Way, IPG, NEESGrid, BIRN, GEONGrid and LEAD, he envisions that data will increasingly become the most demanding challenge. He described progress to this, through Web-form interfaces, and the fusion of data management and interoperable workflow tools. He anticipates building LIGHTWEIGHT Grid infrastructures and a “Google” for Grid data and application services discovery; challenges with scalability of Grid services; and the virtualization of data storage and computational resources. Are Web services really the right foundation for building Grids? Or even for building the galactic-scale Grid operating system?
The panel on “Grids in five years” was another highlight of the conference. According to Fran Berman, we will see more and more innovative Grid-based applications from the commercial sector, which integrate across scale — like e-Bay's Grid-based shopping, real-time multi-media applications, and smart clothes, cars and highways. She also envisions Grids that will be easy to use, secure and efficient, supported by a large number of tools for performance analysis, optimization, debugging, administration and more. Mark Linesch from GGF described the transition from today's “old” world of static, silo, physical, manual, application-oriented IT, toward the “new” dynamic, shared, virtual, automated and service-oriented world. An important step toward success is the recently announced closer collaboration of GGF with the Enterprise Grid Alliance on standards and their implementations in enterprise environments. Carl Kesselman described the future service-oriented architecture, decomposable and dynamically integrating, with applications as services composed into workflows; on-demand provisioning of resources delivered to services; and the creation virtual communities with the challenge to support dynamic policies and trust, understand what policies will work (and what not), and provide quality of services and agreement mechanisms. Finally, Fabrizio Gagliardi described the evolution from mainframes to mainstream personal and departmental clusters becoming powerful nodes in a Grid — a vision obviously affected by his recent move to Microsoft. Watch out and listen carefully to Bill Gates' keynote at SC'05: we will hear more from Microsoft in the near future!
By the way … after six successful years and growing number of participants, the oversized Grid workshop will divorce from the annual Supercomputing conference next year and become a standalone event — the 7th IEEE/ACM International Conference on Grid Computing. It will be co-locating with the 2006 Cluster Conference, which both will take place in September in Barcelona, right after the famous annual Fiesta weekend.
About Wolfgang Gentzsch
Wolfgang is currently the coordinator of D-Grid, the German five-year grid initiative which aims at providing a persistent e-science infrastructure for the German academic and research community. He is also an area director of major Grid projects at the GGF Steering Group. Last, but not least, he is a visiting scientist at RENCI Renaissance Computing Institute at UNC Chapel Hill. In this role, he is conducting a study on a next-generation IT infrastructure for the North Carolina higher education and research community.