ISPIDER

1
(No Transcript)
2
Resource Integration

• OGSA-DQP not fully deployed
• Fixed a number of bugs
• Still unable to access parts of resources
• Tasks
• Fix OGSA-DQP problems
• Adjust OGSA-DQP related AutoMed code to work with
  new AutoMed query processor
• Deploy canned queries service over integrated
  resource
• Use Manchester-based PRIDE and PepSeeker
• Integrate service with ISPIDER Central

3
Query Optimisation

• Initial testing of the integrated resource
  indicated query performance issues
• Developed optimisers
• Equijoin elimination optimiser
• Unnesting optimiser
• Nesting optimiser

4
Query Optimisation

• Equijoin elimination optimiser
• Resource identifiers are disjoint
• Equijoins on identifiers will therefore produce
  no results after retrieving possibly large
  amounts of data from resources.
• Optimiser examines reformulated query and
  eliminates such equijoins

5
Query Optimisation

• Unnesting optimiser
• After query reformulation, query is usually
  nested
• If nesting contains constructs from multiple
  resources, it cannot be pushed down to the
  resource as a whole
• DQP can handle a single level of nesting
• Unnesting optimiser helps alleviate part of this
  problem

6
Query Optimisation

• Nesting optimiser
• Sending multiple separate small queries to each
  resource is ineffective
• Cheaper to create few large queries for each
  resource
• Nesting optimiser identifies and groups
  subqueries designated for a single resource

7
Resource Evolution

• Resource schemas constantly evolve
• Global schema may evolve
• AutoMed handles local and global schema evolution
• However, changes are currently handled manually
• Implementing evolution package for AutoMed
• Package will help the efficient maintenance of
  the integrated resource

8
Resource Evolution

• Example of local schema evolution
• Define the evolution of the local schema as a
  transformation pathway from the old to the new
  local schema
• In many cases, integration network is repaired
  automatically

9
Resource Integration

• Current integration is Global-As-View, i.e. have
  defined global schema constructs in terms of
  local schema constructs.
• Reason
• GAV query processing easier and already
  implemented in AutoMed

10
Resource Integration

• When local resources evolve frequently,
  Local-As-View integration is preferable, i.e.
  define local schema constructs in terms of
  global schema constructs
• AutoMed team developed and implemented LAV query
  processing

11
Future Work

• AutoMed DQP ISPIDER Central integration
• AutoMed performance and usability evaluation
• Service integration with AutoMed and Taverna
• Schema evolution package