Evaluation of UDDI for the UK eScience GRID

1
Evaluation of UDDI for the UK e-Science GRID

• Matthew J. Dovey
• Technical Manager
• Oxford University e-Science Centre
• http//www.e-science.ox.ac.uk
• matthew.dovey_at_oucs.ox.ac.uk

2
UDDI

• Universal Description Discovery and Integration
• Provides data model for describing businesses and
  services
• Provides API for accessing registries
• UBR Public Registry maintained by IBM,
  Microsoft, HP etc.
• But UDDI can be used for private/community
  registries

3
UDDI History

• Initial simple specification to encourage take-up
  and early implementations
• Initial rapid turn-around of new versions
• Version 1.0 late 2000
• Version 2.0 mid 2001
• Version 3.0 mid 2003
• 2000 - 2002 lead by UDDI Consortium
• 2002 - moved to OASIS TC
• Present activities
• Best practices/technical notes
• V.Next discussions

4
UDDI Data Structure
V2
5
UDDI API

• Security
• get_AuthToken, discard_AuthToken
• Publication
• save_ (create, update), delete_
• Inquiry
• find_, get_
• template based queries
• No relevance/ranking
• Negotiation (QoS, price etc.) should happen
  outside of UDDI (but UDDI may contain endpoints
  to such negotiation services)

6
UDDI v3

• Subscription
• Synchronous
• Asynchronous (via callback webservice)
• Digital signature support for authenticating
  provenance
• Custody transfer
• Explicit node replication API
• V2 left to implementation
• Migration of data between registries
• UBR as registry of key generators
• UDDI Policy modeling
• UDDI Extensibility
• All UDDI objects can be referenced by http gets
• Additional query modifiers, category groups,
  internationalization, etc.

7
UDDI v.next

• Compatibility changes for SAML, WS-A, WS-I,
  WS-Policy, BPEL, etc.
• Better external taxonomy support (OWL) etc.
• Better searching (e.g. range searching, semantic
  searching)
• More granular access control (by role, entity,
  action)
• Life of data (stale data)
• Trustworthiness (integration of trust and
  identity services)
• Federation (representing registries within
  registries)
• Different comparisons for category groups

8
ETF Evaluation

• UK e-Science GRID Engineering Taskforce (ETF)
  evaluates technologie to underpin the UK
  e-Science infrastructure and the National Grid
  Service (NGS)
• The Service Registries Workpackage was to
  evaluate UDDI for a services registry for the UK
  e-Science Programme which was
• geographically distributed
• load-balanced
• fully redundant/replicated
• The evaluation work took place between June 2004
  and December 2004
• UDDI nodes deployed at four locations
• Daresbury
• NeSC
• OeSC
• WeSC
• ebXML Registry considered although no evaluation
  work yet undertaken

9
ebXML and the Semantic GRID Conceptspresented
with Carl Mattocks at Semantics in GRID and P2P
Computing at WWW Conference 2004
10
UDDI v2 versus v3

• UDDI v3 offers
• Replication API (for load balancing and
  redundancy)
• Subscription API (for partial replication between
  e-Science registry and private registries)
• PKI Signing (for provenance and trust)
• However
• Few available commercial UDDI v3 offerings
• Systinet
• Computer Associates
• No open source UDDI v3 offerings
• Apache jUDDI (version 2 with some v3 plans)
• Novell NSure UDDI registry (version 2, no dates
  for v3)

11
Evaluation Stages

• Based on OpenSource tools
• Installed jUDDI at the four sites
• Implemented PKI authentication
• Investigated clients
• UDDIBrowser http//www.uddibrowser.org
• Did some additional work on publishing, donated
  to the UDDIBrowser source
• CCLRC web based client
• Customised to support additional contact
  extensions
• Apache Cocoon based client
• Identified test users
• Developed cookbook based on user feedback
• Finding a webservice
• Publishing a webservice
• Implemented replication
• Database level replication (mySql, Postgres)
• Discovered open source offer quantity of
  solutions rather than quality
• Most based on master-slave rather than peer
  replication (i.e. load balancing/redudancy of
  reads but not writes)

12
UDDI TC/UK e-Science Workshop

• Oxford, September 2004 after UDDI Technical
  Committee Face to Face
• Meeting to discuss e-Science/GRID requirements of
  V.Next
• Issues discussed
• Access Control Lists
• More granular to allow others to add metadata to
  services
• Replication allowing peer to peer editing (rather
  than edit only at owning node)
• Will form part of next version
• Quality of Service
• To express availability, reservation, leasing
  etc.
• Unclear whether this is a role of a registry or
  registry locates services which can express this
  information themselves (e.g. via WSDM)
• Searching
• Better searching facilities (e.g. ranges)
• Will form part of next version including OWL
  based semantic searching
• Stale data
• Including short lived data
• Will form part of next version
• Authentication
• Need to support additional mechanisms
  (Shibboleth, PKI)
• Contacts

13
Conclusions

• UDDI can provide an infrastructure for e-Science
  but
• Current OpenSouce implementations still in
  development and poor support for v3
• Neither v3 nor database level replication
  supports peer replication, so load balancing and
  redundancy limited to searching not editing
• Poor ACL Granularity
• Poor edit provenance tracking
• Poor support for short lived data
• Poor search model for semantic and range
  searching
• Poor support for external contact directories
• Issues 2 7 being addressed by UDDI v.next work
• Issue 1 should improve as Apache jUDDI and Novell
  UDDI Server develop
• Some issues (1, 4, 6) being addressed in the OMII
  Grimoires Registry project

14
Links

• UK e-Science GRID Engineering Task Force
• http//tyne.dl.ac.uk/ETF/index.shtml
• OASIS UDDI Technical Committee
• http//www.oasis-open.org/committees/tc_home.php?w
  g_abbrevuddi-spec
• Grimoires Registry Project
• http//www.omii.ac.uk/mp/mp_grimoires.htm