State of Play: The Annual DSG Review

1
State of Play The Annual DSG Review

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

2
Outline

• Introduction.
• DSG Activities
• OGSA Testbed,
• MPJ,
• GridRM,
• JEM
• UISB,
• jGMA,
• Others.
• Group Performance.
• SWOT.
• Summary.

3
Introduction

• Aims of the talk are
• Present the annual review of the DSG, what has
  been done and where we are going
• Have a look at some of the strengths and
  weaknesses of the group.
• Cross reference what we said we would do and what
  we have done!
• Try to take a wider view of future research and
  where we will be five years downstream!

4
The Distributed Systems Group

• The Last Years Activities

5
OGSA Testbed

• One year project Sep 2003 Oct 2004.
• With Daresbury, Manchester, Reading and
  Westminster.
• Aimed to install GTx at all sites and deploy apps
  across testbed.
• Hong (funded) and Mat worked on the effort.
• Regular testbed meetings at Westminster.
• Fund with Starbug firewalls

6
OGSA Testbed - Outcomes

• Good Aspects
• Knowledge and learning
• Hong at GoSC event today,
• SC Demos (jGMA GridRM),
• Papers at AHM (3 poster).
• Integration with the wider e-Science programme,
• Status and profile e-Science, Globus Alliance
• Useful platform for testing over the learning
  curve!,
• Leverage for future collaborations and proposals,
• Even though funding over, testbed partners want
  to continue working together.
• Won Web site bronze prize at AHM!
• Oh and the case of wine

7
OGSA Testbed - Outcomes

• Not such good aspects
• Painful process, working with less than stable
  software,
• Hiccup and change of direction in Jan 04,
• What did the other testbed actually do?
• No following on funding
• GT4 effort within ETF we only have a watching
  brief!
• Firewalls remember GridFTP!
• Where to place our efforts with this, in the
  future.

8
MPJ Progress and Plans

• Background
• Want to produce a reference Java messaging system
  MPJ.
• Aamir started working on the effort in Nov 2003.
• Working with Bryan Carpenter now at OMII in
  Southampton.
• MPJ layered architecture
• Message passing API similar to MPI.
• Secure runtime infrastructure.
• Current status
• NIO device complete,
• Sub-set of API,
• Basic infrastructure embryonic.
• Issues
• Technical issues with NIO,
• Moving goal posts emphasis on
  API/Infrastructure.
• Future
• Complete reference implementation,
• Java MP/SM programming framework using MPJ/JOMP.

9
Resource Monitoring

• Progress
• GridRM framework has made significant progress
  over the last year
• Consists of two layers local and global (linked
  via jGMA).
• Deployed GridRM over international testbed.
• On-line demos - Cluster 2003/Taiwan/AHM
• Garry has submitted his PhD dissertation,
• RM review for Core Technologies book.

10
Resource Monitoring
SDSC
11
Resource Monitoring

• Future
• Garry from 8th Nov - will take up RA post,
• NeSC SC04 demo,
• SPE journal paper from thesis!
• Basis for e-Science proposal,
• Development
• Enhance/extend testbed,
• More drivers,
• Installation/packaging,
• Sub-projects
• UG working on Portal,
• Afrasyab Bashir schedulers broker.

12
JEM

• JEM (recap) consists of
• Software modules that provide OS and other
  services,
• A mechanism to put them together and run them on
  the designated small device.
• Core modules are complete JEM can run simple
  services on a VM on Linux or Myrinet NICs.
• The JEM project was completed in January 2003
• Hong submitted and successfully defended his PhD
  thesis.
• JEM now in limbo, needs
• Hardening,
• Soak testing,
• Packaging and release plus decent Web pages!
• Further development like a Java VM, modules,
• Papers

13
UISB - 1

• The UISB project is developing technologies to
  interact, search and download Information
  Services data, e.g. UDDI, LDAP, Jini LUS.
• Underlying idea
• Search IS and retrieve all manner of high and
  low-level information that can be examined and
  potentially used by a developer
• Textual human readable view,
• Technical so can potentially create client to
  interact with the found service.
• Technologies
• Using Eclipse as the development/deployment
  platform.
• Semantic Web technologies
• RDF as unified data store so can join disparate
  info.
• XML/XLST/Xpath/Xquery/etc

14
UISB - 2

• Progress
• Slowed by Eclipse development environment!
• Interaction with UDDI servers
• For details see Richs talk.
• Looked at MDS3/4 problems
• Lack of documentation,
• Changing APIs (CoG),
• Unstable code.
• Currently looking at Jini (Java objects).
• Near future
• Break out components from Eclipse,
• Look at using the RDF store with Mark Hs
  project,
• Integrate components into a Web portal using
  Portlet technologies.
• Rich giving a talk in Italy in December.

15
jGMA

• Wanted a Java-based, lightweight!, easy to use,
  extensible version of GMA for GridRM.
• Was originally called JavaGMA, now jGMA.
• Testing and benchmarking over first half of 2004.
• Paper accepted and presented
• At UK LUG in Bournemouth January 2004,
• At AHM, Nottingham August 2004.
• Started of with blocking/non-blocking API first
  alpha (binary) release in June.
• Refactored in early summer
• Now non-blocking API
• Updated end-point addressing.
• Blocking API as add on later.

16
jGMA

• Progress
• Non-blocking jGMA API has been thoroughly tested,
• Leasing, partially implemented (found very
  important),
• Demo for SC 04 looks neat.
• Working on a scalable virtual registry
• Based on persistent XML/SQL/text databases/files,
  
• Being designed Mat will describe this work in
  detail in his seminar.
• Further benchmarking compare with PyGMA/R-GMA.
• Full release.
• Full integration within GridRM,
• Start actively using it in other systems
  on-line Games!

17
OGSA-DAI 1

• Initiated within the OGSA Testbed project.
• Working with Southampton on GITS for OGSA-based
  systems
• Perl scripts to examine and show the status of
  Grid services.
• Interested in OGSA-DAI as a pervasive technology
  for integrating relational databases.
• Decided to build a prototype system that pushes
  GITS data into a database that can then be
  searched and manipulated using an OGSA-DAI Grid
  Service and viewed via a Web browser
• Helen is working on this project.

18
OGSA-DAI 2

• Progress
• Perl script to parse GITS completed,
• Poster in AHM,
• A number of technical problems
• OGSA-DAI not rock solid yet,
• Interaction with mySQL and Xindice problematic.
• Starting to overcome hurdles.
• First prototype expected to be up and running by
  the end of November!
• Initiate a discussion with Bob N. to talk about
  using OGSA-DAI with SDSS look at intelligent
  and efficient search mechanisms on large
  distributed databases
• Research and develop a number of tools and
  techniques.

19
SDSS
20
Other Projects

• Grid Performance
• International effort
• JISC funds to help with the workshop and pay for
  a report.
• Workshop held at Imperial College in May.
• Final report submitted to JISC in September 04.
• Potential follow up events maybe NeSC!
• Portal development (Hong)
• Part of a JISC proposal to develop a bunch of
  services for Sakai.
• GridSphere being used for the liquid Crystal work
  (OGSA testbed).

21
Portal Effort
22
Performance Metrics

• Need to look at what are the measures of success
  for the group to be able to assess the potential
  of the DSG.
• The RAE uses
• PhD completions,
• Income,
• Papers,
• Esteem/activities.
• Are these good metrics?

23
Growth of DSG
Helen Mark H.
24
DSG Papers (per year)
X - Conferences workshop
X - Int. Journal
5
4
Publications
2
1
98
99
00
01
02
03
05
04
Year
25
Funding proposals

• 2002 wrote/contributed to 7 proposals
• 2 funded!
• 2003 - wrote/contributed to 8 proposals
• 2 funded!
• 2004 - wrote/contributed to 5 proposals
• JISC JCSR VRE PD for two years, with Daresbury,
  Lancaster, Oxford.
• Basic technology programme - Paradigm Shifted
  Virtual Screening PD for 3 year, with Bio.
• Core e-Science Grid performance modelling PhD
  for three year, with Warwick and Nottingham.

26
DSG Income
EPSRC 45K
50
Myri
CLRC 14K
40
DERA 36K
EPSRC E-science - 80K
Money (k)
DERA 36K
20
IBM 14K
JISC JCSR 20K
NHSE 5K
10
TOPIC 10K
TOPIC 10K
CLRC 10K
TOPIC 10K
CLRC 4K
CLRC 4K
CLRC 4K
98
01
99
00
02
03
04
05
Year
27
DSG Membership

• MAB staff.
• Garry
• Submitted PhD Oct 2004,
• RA from Nov 8th 2004,
• Take AstroGrid post in Feb 2005.
• Hong
• Submitted and defended PhD Feb 2004,
• 2-year RA post
• OGSA Testbed RA (1 year), Faculty RA (1 year).
• Rich - Faculty PhD bursary (1 year to go).
• Mat - DTA PhD bursary (2 years to go).
• Aamir - DSG PhD bursary (2 years to go).
• Helen CSSE bursary (2 years to go).

28
Esteem (reflects on the group)

• TCSC (co-chair) and TCPP (exeCom).
• Visiting chair Westminster/Reading.
• Steering Committee of Cluster/CCGrid/Grid.
• DSO section editor.
• Invited talks.
• Grid Book early 2005.
• Reviewing (books, journals, conferences).
• Membership of conference PCs.
• Organisation of events.
• NeSC events.
• Rapporteur for IST CFP, Nov 2003.
• Hits on, and prize for Web site

29
DSG Presence
30
DSG Activities
31
DSG Activities

• ICCC
• Melbourne, Australia, Dec 1999.
• Chemnitz, Germany, Nov 2000.
• Newport Beach, California, Oct 2001.
• Chicago, Sep 2002.
• Hong Kong, Dec 2003
• San Diego, Sep 2004
• Boston, Sep 2005
• Barcelona Sep 2006
• Texas 2007
• CCGrid
• Berlin, Germany, May 2002, Tokyo, Japan, May
  2003, Chicago, April 2004, and Cardiff 2005
• Grid (with SC)
• Denver 2001, Baltimore 2002, Phoenix 2003,
  Pittsburgh 2004,
• With SC in 2005 too.

32
DSG Metrics

• PhD completions
• Hong - February
• Garry November!
• Papers need to produce at least one conference
  paper per year and one journal paper per PhD
  completion.
• Income average, hopeful of success over coming
  year.
• Esteem group is well know in the US and has
  started to establish a better reputation in the
  UK and EU.

33
SWOT Analysis

• Strengths
• Emerging national recognition (Testbed, AHM, SC
  demos),
• Very good relationships with Daresbury and via
  OGSA Testbed.
• Partially grown group via visitor and CSSE,
• Semi-autonomous (good and bad).
• Group has extensive and complimentary skills,
• Continue to have greater interaction with the ICG
  and growing these links especially now Bobs
  around.
• Well connected, building even stronger links with
  EU and UK groups.
• Knowledge in depth Java technologies, gather
  experience with WS, and various W3C XML
  technologies,
• Profile via TCSC, DSO, and DSG Web presence,
• Undertaking synergetic projects which leads to
  good group focus.

34
SWOT Analysis

• Weaknesses
• UoP CS research is still low-profile, making it
  doubly hard to get funds.
• Still no internal UoP staff help with getting
  funding.
• Still trying to get other UoP staff involved in
  DSG research, potential with ICG and CSSE PhD
  student.
• Administration and administrative effort is
  challenging and time consuming, Chris Duncan is
  helping us no end, but it still sucks a lot time.
• VERY Ham-strung having to work through ISO, e.g.
  fire walls, general attitude, etc.
• Still little industrial and/or commercial
  connections.
• May be losing Hong hard to replace!
• Semi autonomous justify our existence, not a
  very large umbrella.

35
SWOT Analysis

• Opportunities
• OGSA Testbed Great opportunity which has long
  term
• TCSC, educational and other collaborative
  changes,
• Mark Holliday various openings,
• Building up links with Biology (Martyn Ford
  BTP),
• Strong national links
• Daresbury, Lancaster, Westminster, Reading,
  Southampton, and Manchester.
• Grid book funding for companion Web site and
  development of material.
• Getting more involved in the e-Science programme,
• Semi-autonomous,
• With ICG we have the best research culture in the
  UoP!
• IEEE CS TCSC/DSO

36
SWOT Analysis

• Threats
• FUNDING
• Involvement with GPN,
• Recruitment of students, RAs and more senior
  staff.
• Semi-independence via the ICG!!
• Internal faculty funding will our funds be
  there in five years!
• Getting bogged down in teaching and
  administration a large overhead is associated.
• Making sure DSG income gets tagged for the DSG
  research.
• Not retaining good staff! losing staff!

37
SWOT Conclusions.

• As usual interesting comparing this SWOT with
  that of last year.
• Some fears regarding threats and weakness have
  happened
• Group has not grown limited funding, but had
  additional input from Helen and Mark H.
• Teaching and university processes are sinks of
  time and effort!
• With respect to our strengths and opportunities
  we seem to be taking advantage of them
• Our reputation, skills and presence is helping us
  a lot,
• Optimistic about funding proposals.
• I still think our perceived strengths and
  opportunities out-weigh the weaknesses and
  threats, but we still need to be careful!
• The biggest threats and weakness are on the still
  on funding side (internal and external).
• Our strengths are focussed on applied research,
  useful skills, good interactions, and lots of
  enthusiasm!

38
Summary

• I think weve had a reasonable year.
• We have managed to
• Enhance our reputation,
• Produce some good ideas,
• Continue to get some income,
• Maintaining productive research culture.
• Widening our experience and collaborations.
• However
• No new students this year,
• Limited new funding,
• Lots of doing, probably not enough papers.
• Congrats to Hong successful defence of PhD.
• Good luck to Garry submitted PhD, viva soon!
• I told you we were getting better at Web design.

39
Five Years Hence (2002 view)

• Group size of ten plus permanent staff, post
  docs and post grads.
• Coherent and strongly focused research.
• Independent group working in a strong research
  culture.
• Broken down the external view of research at UoP.
  
• Internationally recognised in a number of related
  fields
• Wide-area resource monitoring,
• Java micro-kernel work,
• Java-based Grid technologies.
• Deployed and regularly downloaded software.
• Papers 30 conference and journal papers per
  year.
• Invites to collaborate/talk at international
  events/institutions.