Service and Resource Matchmaking in ASG

1
Service and Resource Matchmaking in ASG

• Kashif Iqbal, Stefan Decker
• kashif.iqbal, stefan.decker_at_deri.org

2
Outline

• Grid System Taxonomy
• Services in ASG
• Matchmaking Problem
• Requirements
• Components
• Resource/Service Specification languages
• Existing approaches
• OMM, WSMX and myGrid
• Distributed matchmaking
• Discussion and Conclusion

3
Grid System Taxonomy

• ASG
• On demand Service Grid?
• Collaborative Service Grid?

4
Services in ASG

• Services in ASG
• Resource Services
• Core Services (middleware)
• Application specific services
• Service Matchmaking in ASG
• Service specification (service type) matching
• MDS, UDDI, RDF stores
• Service implementation (instance matching)
• JINI LUS
• Resource service matching
• Composed service matching
• Workflows

5
(No Transcript)
6
Example Condor Matchmaker
TypeJob Owneru2 Constraint
other.TypeMachine Memorygt500
OpSysSolaris251 Rankother.Memory
Machinem2
Matchmaker
Resource Requesters (users/programs)
TypeMachine Namem2 Memory1000
OpSysSolaris251 Grp1u1,u2
Constraintmember(other.Owner, Grp1)
TypeMachine Namem1 Memory1500
OpSysRedhat7.3 Grp1u1,u2
Constraintmember(other.Owner, Grp1)
7
Requirements for Service Specification language

• FR
• Data semantics
• Functional semantics
• Execution semantics
• NFR
• QoS semantics (execution time, cost, bandwidth)
• I/O Communities

8
Request/Resource Specification languages

• JDL
• JDML
• ClassAds
• Globus RSL
• XML and XPath
• GridML
• Specialized form of XML representation describing
  physical and observable char of resources
  networks
• RDF (RDF-S) and Triple
• DAML-OIL (OWL)
• DAML-S (OWL-S)
• WSMO/WSML
• RedLine (constraint specification language)
• ACDL/ITL (Concept language)

9
Matching Engine requirements

• Support flexible semantic matching
• Request should have some control on the
  flexibility of the matching
• Minimize false positives and false negatives
• Matching engine should encourage requesters and
  providers to be precise with their descriptions
• Efficiency/Performance
• Integrity checks
• Bilateral constraints

10
Requirements II

• Bilateral matching
• Multi lateral matching
• Set matching
• Gang matching
• Symmetric vs. asymmetric matching
• Expressiveness
• Flexibility and Extensibility
• Service and workflow discovery
• Plan base vs. Registry approach

11
Matchmaker Components

• Filters
• Context matching
• Syntactic matching
• Semantic matching
• Registry
• Workflows, SP registry, Service specifications
• Ontology management
• versioning
• Knowledge base
• Store domain background knowledge

12
Resource Discovery
13
Grid RMS
14
Grid RMS
15
Existing Matchmaking

• my Grid
• Condor/Condor-G
• Central broker
• Constraint language approach
• OMM
• InfoSleuth
• Larks (lack rule based approach for matchmaking)
• WSMO/WSMX matchmaking
• Hawkeye Matchmaking
• Meteor-S (Bottom up approach)
• Registry Ontology
• Ontology at web services level
• Community Scheduler

16
OMM
Advertisements/ Requests
Errors/Reply

• - Matching Criteria (e.g.,
• Request.OS must be
• compatible w/ Resource.OS)
• - Matchmaking Algorithm
• (e.g., return the highest rank
• compatible resource)
• Integrity Checking (e.g.,
• check for consistency)

Ontology-based Matchmaker
Background knowledge About the domain, e.g.,
SunOS, Linux are types of Unix
operatingSystem
Domain Models objects, their properties
relationships between objects (e.g.,
ComputerSystem, OperatingSystem, RunningOS)
Domain Ontologies (Resources,Policies,Requests)
Deductive Database System (Triple/XSB)
17
Distributed Matchmaker

• Interconnect matchmakers in P2P fashion
• Avoid central infrastructure
• Avoid bottleneck and single point of failure
• Share costs
• Example
• Edutella (based on JXTA)

18
Challenges

• Query and Answer Routing
• Flooding kills the network (see Gnutella)
• P2P and Grid
• Integrate P2P capabilities in Globus Toolkit
• Security

19
Matchmaking in WSMO/WSMX/WSML

• Based on matching of logical Goals with WS
  Capabilities
• Goals and capabilities have post conditions and
  effects.
• Capabilities additionally have preconditions and
  assumptions
• WSMX adds concept of conditional Web Service to
  capability

WSMO Registry
WSMX Matchmaker
Step 1
Step 2
Match requester
Goal
Possible Matches
Collection of WS
Step 4
Step 3
Network
ConditionalWS1
ConditionalWS2
20
myGrid Services and Middleware
21
Discussion

• IDL (WSDL) provide binding information
• Matchmaking vs. Claiming
• Claiming
• Search for service providers to bind
  specification to concrete implementations
• SLA, job monitoring
• GRAM (WS-Agreement)
• Resource reservation (Planning)
• QoS (admission control and policing)
• Hard vs. Soft QoS

22
Brokering vs. matchmaking
23
C-2 Intracomponent Relationship
Distributed ASG Registry
Workflow Planner
User
Reasoning Engine
C-3
24
Conclusions

• Matchmaker based on existing technologies
• Semantic based matching
• Distributed matchmaking based on P2P
• Broker vs. matchmaker
• P2P registries vs. Federated approach
• Matchmaker Registry
• Integrated or Isolated
• Registry or registries
• Issues involved
• Redundancy
• Performance
• Use cases for matchmaker

25
Questions/Suggestions

• Thank You!

26
References

• The myGrid project services, architecture and
  demonstrator, Carole Goble, Chris Wroe, Robert
  Stevens and the myGrid consortium, EPSRC
  e-Science Pilot Project myGrid http//www.mygrid.o
  rg.uk
• A Taxonomy and Survey of Grid Resource Management
  Systems, Klaus Krauter1, Rajkumar Buyya2, and
  Muthucumaru