The Great Academia/Industry Grid Debate

1. The Great Academia/Industry Grid Debate November 17 2003 Geoffrey FoxCommunity Grids LaboratoryIndiana University gcf@indiana.edu

2. Topics to Discuss • 1) What will make Grids a Reality (with a big "G")? If it requires "deploying core services with tolerable reliability in enough places", what does the this mean really mean and how do we get there? • 2) What grid products would users really like to see from vendors that would make their grid efforts easier? (What commercial-off-the-shelf grid products could you piggy-back on?) • 3) What does industry really need to see standardized (specifics beyond just simply OGSA) in order to build grid products that can brought to market? I’m academia here • 4) What has to happen to make grid systems easier to install and use (e.g., "Grids for Dummies")? • 5) What is the real scalability and performance of grid information systems? Is it good enough and scalable enough? • 6) How will licensing issues affect how commercial vendors can deploy grid products?

3. What will make Grids a Reality (with a big "G")? If it requires "deploying core services with tolerable reliability in enough places", what does the this really mean and how do we get there? • Maybe a big R for Reality is more important than a big G for Grids • The requirements of • Particle Physics Experiment at CERN • Grid for campus (enterprise) computing or • Teragrid or DoE Science Grid • Is very different from Grid needed • Support K-12 (Education) class • Support e-Science with a Grid linking say 3 geographically distributed research groups with modest computing sophistication • Need simplicity, low cost of entry and low risk that custom services will need major changes as Grid architectures, standards and implementations inevitably change

4. What grid products would users really like to see from vendors that would make their grid efforts easier? (What commercial-off-the-shelf grid products could you piggy-back on?) • I did a Gap Analysis with David Walker – interviewing 100 people and finding many Grid gaps to fill – see http://www.nesc.ac.uk/technical_papers/UKeS-2003-01/index.html • Not many people thought Industry would fill the Gaps • Specific services or architectural components are most likely to be used (as opposed to monlithic solutions) • Admit likely to be much greater than1 Grid and address interoperability/federation of multiple deployments • Also tools enabling service construction/meta-data/Grid deployments (services, security) • Core technology open source to lower risk

5. What has to happen to make grid systems easier to install and use (e.g., "Grids for Dummies")? • The option of much simpler architectures like Gnutella Jini or JXTA • which could be peer-to-peer (as in examples) but • should support pure Web services • Complex Grids (such as that demanded for campuses, CERN data analysis) should have some way of attaching simpler Grids (e.g. 3 distributed physicists discussing signature of a new particle) where simpler Grid not “infected by complexity of central Grid” • Federation again

6. What is the real scalability and performance of grid information systems? Is it good enough and scalable enough? • Even worse we have at least different styles of meta-data • Stored in distributed fashion among the services (as in SDE’s and many portals) • Stored in meta-data catalogs separately from the resource/service/data described • Stored already in non Grid standard compliant application specific meta-data catalog which may or may not also contain data • Need some best practice that can integrate these approaches

7. How will licensing issues affect how commercial vendors can deploy grid products? • Core technology should have a good open source implementation • All interfaces should be open • High value core implementations and “particular non-core services” can be proprietary