OGSA Grid Testbed

1
OGSA Grid Testbed

• Dr Mark Baker
• Distributed Systems Group
• University of Portsmouth
• http//dsg.port.ac.uk/mab/

2
Outline

• Testbed Overview.
• Strategy.
• Progress
• Overall,
• Partners and associates.
• Summary and Conclusions.

3
Testbed Overview

• Partnership between the universities of
  Manchester, Portsmouth, Reading, and Westminster,
  and Daresbury Laboratory.
• One year of funding about half an FTE at each
  site.
• Funded to be in the vanguard of testing and
  experiencing emerging OGSA-based grid
  middleware with the aim of feeding back what we
  learn into the e-Science community.
• Each partner porting applications to OGSA-based
  middleware testbed.

4
Testbed Strategy

• Original strategy was tightly coupled with OGSI,
  this was revised in early February in light of
  the announcement of WSRF.
• Updated strategy is based on gaining as much
  knowledge and experience as possible in order to
  provide extensive feedback to the e-Science
  community.
• In particular
• Testbed Platform follow GT3 ? GT4 (WSRF)
• Applications multiple paths ending up with
  WSRF
• Follow rollout of GT3 ? GT4,
• Web Services GT2.4 ? GT4,
• OGSILite ? WSRFLite.

5
Progress Overall

• Testbed Platforms
• GT3.2 installed, tested and running at all sites,
• Cross-site logins set up and being used.
• Feedback bug/features to Globus developers.
• Regular discussions and meetings.
• Project Web site
• Secure pages via Grid Site.
• Producing and publishing various documents.
• Numerous report papers submitted to AHM.
• All partners working on porting their
  applications with aim of deploying them across
  the testbed.
• Interacting with parallel OGSA testbed.

6
Progress - Daresbury

• Extensive tests of OGSA-DAI and GT3 MMJFS.
• Produced a MMJFS scalability and throughput
  report.
• Deployed a GT3.2 index service application with
  UDDI
• See http//thames.dl.ac.uk8080/index-service/
• Also implemented as in portlet and deployed as a
  portal using Jetspeed.
• Deploying OGSA-DAI as a test service for the NGS.

7
Progress - Manchester

• Announcement of WSRF means that porting
  applications to GT3 is no longer seen as
  strategic.
• Instead, using OGSILite to track OGSI ? WSRF
  evolution - Perl-based implementation of OGSI
• Basis for first open-source implementation of
  draft WSRF specifications.
• Emphasis on interoperability
• Stephen Pickles is leading the Experiences
  document that will be produced by the OGSI
  Working Group of GGF
• Mark Mc Keown participated in WS-RF
  interoperability fest hosted by IBM (19-20 April
  2004, North Carolina)
• Seeking alternative funding to bring WSRFLite
  to production quality

8
Progress - Portsmouth

• Extensive testing and installation of GT3.
• Completed a comprehensive document detailing
  every step in the installation, set up and test
  of GT3 (a moving target!).
• Helping Southampton and Manchester with GITS
  development and testing for OGSA.
• Developing OGSA-DAI GITS service.
• Currently porting Liquid Crystal Portal to
  GridSphere (utilising Portlet technologies) using
  Web Services and GT2.4.

9
Progress Reading

• Later start bootstrapping now.
• Working on porting the Genome related
  phylogenetic tree and Air Pollution codes to run
  on P-GRADE and OGSA respectively (Web Services
  GT2.4).

10
Progress Westminster

• The UoW is exploring the most appropriate way to
  deploying legacy codes on GT3.
• They have a system, GEMLCA (Grid Execution
  Management for Legacy Code Architecture) that
  will allow the execution of legacy codes through
  GT3.
• GEMLCA has a front-end OGSI Grid service that has
  a number of interfaces to submit and check
  computational jobs, and retrieve results.
• A prototype has been implemented and tested by
  deploying the MadCity traffic simulator.

11
Partners

• Southampton
• GITS script is working with GT3.2.
• Serious inter-site testing is just starting.
• Testing infrastructure developed for "base"
  services - registry, GRAM, RFT - potential
  model for more generalised applications tests.
• SZTAKI Research Institute Budapest
• Refactored the PROVE performance and execution
  visualization system as a GT3 service.
• Designed the GT3 version of P-GRADE.
• Created a P-GRADE portal - first prototype up and
  running.

12
Summary and Conclusions.

• OGSA Testbed project progressing well
• Small project that has created productive
  collaboration,
• Partners have broad base of experience and are
  making good progress.
• Lots of external interest in joining testbed.
• Feedback to ETF and other e-Science bodies.
• Updated Strategy
• Testbed Platform follow GT3 ? GT4 (WSRF).
• Applications multiple paths ending at WSRF
• Follow GT3 ? GT4,
• Web Services GT2.4 ? GT4,
• OGSILite ? WSRFLite.

13
Finally Some Concerns

• Convergence towards WSRF
• Testbed is a short-term project ends in
  September 2004,
• Change in focus in February to account the
  emergence of WSRF,
• Probably will only get our hands on only very
  early versions of the WSRF software.
• Plethora of unratified specifications
  developing applications to moving targets.
• Spending excessive amounts of time installing and
  testing GT releases another moving target!
• Limited code-base, like to see alternatives to GT
  provide greater choice and maybe help increase
  quality through competition.

14
Testbed Web Site
http//dsg.port.ac.uk/projects/ogsa-testbed/