Savas Parastatidis Principal Research Associate School of Computing Science

1. Building Grid applications using Web Services(The Web Services Grid Application Framework)CERN – 4 March 2004 Savas Parastatidis Principal Research Associate School of Computing Science University of Newcastle upon Tyne savas@parastatidis.name http://savas.parastatidis.name

2. Outline • The Grid • Service Orientation and Web Services • WS-GAF • WS-RF • Conclusions & future plans

3. “What is the Grid?” • “Neo, the Grid is everything you would like it to be” • IBM: on-demand computing • HP: utility computing • Microsoft: seamless computing • ORACLE: 10g • Sun: Sun Grid Engine • Intel: Seti@home or whatever makes money • Globus: Distributed computing revisited • … • The Newcastle team: Internet-scale distributed computing

4. The Grid vision • Build applications that span organisations • Create virtual organisations • Seamless integration • Hide use of resources, network, infrastructure • Provide a new computing experience • Enable e-Science, new commercial applications

5. The promise of Web Services • Cross- and intra-organisation integration • Interoperability • Glue for heterogeneous platforms/applications/systems • Standards-based distributed computing • Based on the concepts of Service Orientation

6. Service Orientation • Built around the concepts of service and message • A service is the logical manifestation of some physical resources (like databases, programs, devices, humans, etc.) and/or some application logic that is exposed to the networkand • Service interaction is facilitated by exchanging messages • A service adheres to a contract • Describes the format of the messages exchanged • Defines the message exchange patterns in which a service is prepared to participate

7. Service Orientation • Don Box’s four tenets about Service Orientation • Boundaries are explicit • Services are autonomous • Services share schema and contract, not class • Service compatibility is determined based on policy Source: “A Guide to Developing and Running Connected Systems with Indigo” http://msdn.microsoft.com/Longhorn/understanding/mag/default.aspx?pull=/msdnmag/issues/04/01/Indigo/default.aspx and various talks

8. The Anatomy of a Web Service • Large grained, loosely coupled • Performance, scalability, maintenance, re-use, etc.

9. A Web Service

10. Web Services (SOAP and WSDL)

11. Service-orientation vs Resource-based • Loose-coupling • Encapsulation • Good scalability SOA • Tight-coupling • Easily breakable applications • Poor scalability Resource-based

12. Interactions and State • Stateless interactions allow for loose-coupling • State remains internal to the service (encapsulation) • Applications are built through message exchange patterns and well-defined message formats

13. The “stateless vs statefulness argument” Web Services just “computational entities” or “procedures”? Two types of state • Service internal state Not our concern • Interaction state Contextualisation as the means for message correlation State

14. Web Services Interaction • Stateless interaction Hey, I am savas

15. Web Services Interaction • Stateless interaction Welcome Savas

16. Web Services Interaction • Stateless interaction I would like £100, please

17. Web Services Interaction • Stateless interaction No can’t do Don’t know you

18. context Web Services Interaction • Stateful interaction Hey, I am savas ctx

19. Web Services Interaction • Stateful interaction Welcome Savas ctx

20. Web Services Interaction • Stateful interaction I would like £100, please ctx

21. Web Services Interaction • Stateful interaction Your £100 Have a nice day! ctx

22. WS-GAF

23. Motivation • We believed that OGSI unnecessarily introduced a new mandatory infrastructure layer on top of Web services • which was object-oriented in nature • Our goals • Analyse Grid requirements • Propose an alternative solution based on current WS specifications and practices • Start a technical discussion within GGF • Emphasize the importance of OGSA and its high-level services • Build solutions using WS-I and WS-Security

24. Grid requirements • Stateful interactions • Contextualisation • Resource identification • Metadata • Grid Resource Specification • Lifetime management of resources • Just a high-level service interface • Lifetime information • Part of the metadata WS-Context, WS-Security, WS-Transactions, WS-Coordination, BPEL (message correlation), etc. etc. etc. URN: Uniform Resource Names Just an XML Schema document At the OGSA level

25. Identity • Resources are usually hidden • There are cases where resources need to be identifiable outside an organisation’s boundaries • Grid Resource Identifier (GRI) (like an LSID) • Everlasting, unique resource identifier (Uniform Resource Name, URN) • Can be stored in a database or printed in a journal • Decoupling of identity from interface The resource is identified separately from the interface that can provide access to it

26. Use of identity

27. The Grid Resource Metadata Document • Functionality equivalent to Service Data Elements • Everything implemented using existing technologies and tooling • Not Grid-technology specific (it’s just an XML Schema document)

28. Example: Using a registry

29. From Web services to the Gridand back again

30. The story so far • OGSI was accepted as a specification in June 2003 • Issues with meeting the “interoperability requirements” but everyone ignored them • We published our views in August 2003 • A response was announced by Ian Foster but was never delivered • We presented WS-GAF in GGF9 (October 2003) • End of October 2003 IBM announces to the WSDL working group that they are changing their position on “state” • WS-RF is partially released in January 2004 • It’s pitched as the “convergence point” between Web Services and the Grid

31. Other developments • Microsoft are currently not supporting it • HP are interested in WS-RF for their single-systems management products • Oracle are not keen • Sun? we don’t know • Debate in UK on best route forward • And while the discussions on how to change the underlying infrastructure (again!) are starting, we have decided to build Grid applications with what we have now • But first…

32. WS-Resource Framework - Introduction • A suite of specifications • Half of the specifications were introduced in January • WS-Resource Properties, WS-Notification, WS-Resource Lifetime • WS-Renewable References, WS-Service Groups, WS-Base Faults

33. Initial Impressions • As per our August 2003 document (http://www.neresc.ac.uk/ws-gaf) • A clear separation between the terms “service” and “resource” • Services are not dynamically created; they are deployed • Services provide operations on multiple resources (one-to-many; no implicit one-to-one association between the two) • Factorisation on the functionality into separate specs (not all the way yet though) • A document-based approach to resource properties • Notification as a separate, higher-level service • WSA-friendly specs (no more GWSDL) • Respect to existing tooling; use it without modifications • Still issues with the conceptual model (scalability, loose coupling, is there an actual need for it… use cases?)

34. The conceptual model • Very similar with OGSI • Effectively it’s J2EE or CORBA with angle brackets • Resource identity part of the infrastructure • Stateful interactions only with resources • Applications built through resource sharing rather than message interactions

35. The conceptual model Talking to service instances that have an 1-1 association with a resource Talking to resources

36. Examples (using URNs)

37. Examples (using URNs)

38. Examples (using URNs)

39. Examples (WS-RF)

40. Examples (WS-RF)

41. Examples (WS-RF)

42. Conclusions

43. Benefits of WS-GAF • Simplicity and Minimalism • Meets Grid requirements without inventing new infrastructure • Uses existing contextualisation and addressing specs • Uses URN for resource identification • Low entry and maintenance costs for new Grid services • W3C Web Services Architecture conformant • No changes in the semantics/characteristics • Synergy with existing WS specifications and practices • Avoid effort involved in influencing WS community • Avoid risk associated with not being able to influence WS community • Moves the focus from the low-level “networking” to the more important OGSA platform • OGSA should be the focus for Grid standards

44. Benefits of WS-GAF • Supporting tools • No divergence from WS technologies and tools (e.g., WSDL) • Do not need anything other than existing industry and open-source provided WS tools and platforms (e.g., ASP.NET, Axis, etc.) • Education materials • Industry investment • Performance and scalability • Encourages large granularity, loose coupling (richer, fewer message exchanges) • Explicit identification of resources • Resource lifetimes, everlasting resource IDs • Distributed system technology independence where possible

45. Summary • We believe that WS-GAF meets the requirements for building Grid applications while using today’s WS specifications and practices • We believe that WS-GAF has a range of benefits • We are interested in building on existing WS specifications and utilise middleware that is made available by companies or the open source community (e.g., .NET, Axis, IBM WSDK, Java JAXB, etc.) • What’s next?

46. WS-GAF Application • We need to build large-scale Grid applications using Web Services in order to find out what is actually required • Aims • Define the characteristics of a “typical” Grid application • Demonstrate the applicability of the WS-GAF approach in building Grid applications • Learn from the challenges of constructing a truly global, distributed, scalable, loosely-coupled application • Working on a “typical”, global-scale Grid application with international partners • We encourage everyone’s involvement

47. Searching for “White Dwarfs” • Jim Gray’s SkyServer • Combine info from other databases • Utilise computational resources • Visualise We are looking for these

48. People and Links • Paul Watson (Paul.Watson@newcastle.ac.uk) • Savas Parastatidis (Savas.Parastatidis@newcastle.ac.uk) • Jim Webber (Jim.Webber@newcastle.ac.uk) Web Services Grid Application Framework (WS-GAF)http://www.neresc.ac.uk/ws-gaf ws-gaf@newcastle.ac.uk Join by sending a message to mailbase@newcastle.ac.uk including the following line in the body join ws-gaf YourFirstName YourLastName

49. Thanks • DTI • JISC • UK e-Science Core programme