OOSTech 2005 Summary

Keynote speech by Dan Reed, vice-chancellor for IT at UNC-Chapel Hill and Dir of the Renaissance Computing Institute.

• Benefits of Standards: interoperability, separation of concerns, reuse, independence, dependability, sharing, commonality.
• Prerequisites for web services: metadata standards, standard communication mechanisms, resource discovery and registration
• Mentioned that 40% of eBay listings are now via API calls
• Web services are not a panacea: standards debates can be endless, code is shifted from modules to interfaces
• Danger of "Death by CS Abstration" - all problems can be solved by another level of indirection.

Intro by Phil Bodgen (GoMOOS)

• There is ocean observing. There is need.
• Brief history of OOS Tech. First in 2003. Meet, greet, share ideas, initiate effort to become interoperable.
• Standards enable discovery, integration, multi-use, innovation, & revolution, but require community.
• Goals for 2005: virtual org sharing data; make a new kind of distributed lab for reserach applications; advance science of environmental prediction & hazards planning; enable transition from research to operations.
• The 2005 charge:
  • Hear about current thinking in web services
  • Learn about state of the art (web services 101 and current activities & standards processes)
  • Hear about vision of the future: the grid
  • Decide on path forward: strategy & tactics

Web Services Panel Talks

• Luis Bermudez, Software Developer Liaison of the Marine Metadata Initiative.
• Ilkay Altintas, Asst Drr, Nat'l Laboratory for Advanced Data in San Diego
• Ben Domenico, Deputy Director of the Unidata Program Center.
• Michael Botts, Principal Research Scientist, Earth System Science Laboratory, UA-Huntsville.

They discussed their technology including the use of web services, sensor ML, MMI, Kepler (open-source scientific workflow system that makes use of grid-based approaches to distributed computation)

Web Services 101 by Marc Hadley, senior staff engineer at Sun

• Defined a web service as really just a service on the web (e.g. Google, Amazon), but they are designed for human interation. When we say web service, we're thinking of a service that facilitates machine-to-machine interacion.
• To make a machine friendly web:
• Enable machine-machine interaction
• Don't reinvent the wheel to faciliate web services: the web is the most successful distributed system humans have thus far built.
• Maintain platform and language independence.
• Require semantic clarity in content (XML)
• XML better than HTML, since HTML is concerned with presentation (fonts, paragraphs, order). XML separates data from presentation. XML structure much more rigid than HTML; you can get away with things with HTML - it might look strange, but it will work; not so with XML.
• Security thoughts:
• authentication
• confidentiality
• authorization
• auditing
• integrity
• non-repudiation
• trust, policy
• 2 ways of securing message: secure pipe (e.g. ssl), message layer
• Pros of message layer: endpoint-to-endpoint
• Cons of message layer: relatively new specs, frameworks need profiling (Sun working on this), more moving parts
• Discussed tools for mapping XML to Java classes: JAXP (processing), JAXB (binding), JAX-WS (like JAXB but takes in a WSDL rather than a schema)
• Demonstrated Amazon's API - can access with XML by HTTP or SOAP/WSDL (80% use XML via HTTP)
• Discussed AJaX (Asynchronous Javascript and XML) - e.g. Google maps

Talk by Ian Foster of Argonne National Laboratory

• Discussed WSRF - web services resource framework, or web services for grid computing.
• Discussed importance of state. Few useful services are truly stateless, but WS interfaces alone do not provide built-in support for state.
• WSRF defines conventions for managing state.
• Globus Toolkit v4 (GT4) - public-key authentication, extensible authorization framework based upon web services standard, credential management service, community authorization service, standalone delegation service.
• GridFTP in GT4: 100% Globus code, IPv6 support, 27Gbit/s

Talk by Paul Avery, UFlorida

• The Trillium Grid
• 150 people
• The Virtual Data Toolkit - a unique laboratory for testing, supporting, deploying, packaging, upgrading, and troubleshooting complex sets of software.
• The Open Science Grid - national production quality grid computing infrastructure for large scale science, built and operated by a consortium of U.S. universities and national laboratories. A continuation of Grid3. 50 sites, 15,000 CPUs.

Cyber-Infrastructure Implementations

• Matthew Arrot, Paul Avery, Phil Bogden, Ilya Zaslavsky.
• Panel Q&A:
• Best practices learned from their apps: SCOOP has a wide range of requirements, ocean community is not addressing these; need to stand up and ID which are the most pressing needs & then learn from existing examples in different fields but in the same areas (data grids)

Next Steps:

• Create a clearinghouse site with announcements of test beds, schemas, community activities, documentations, cookbooks, and how-to's, featuring surveys.
• Share solutions between ORION, IOOS, GEON, NEON, SEEK.
• Tie in international efforts.
• OceanUS should work toward convergence of methodologies and provide the role of endorsing
• Small working group volunteered to work with DMAC steering committee; Jeremy was one.