System Functions Deployment in FMG-QCMA

FMG-QCMA is built with a number of system components derived from QCMA.

Those system components existing in QCMA, which include Quali-Fuzy Classifier, UDDI OWL-S, Quali-Fuzy Engine and Quali-Fuzy Moderator, are evolved and replaced by FMQ Distributor, FMQ UDDI OWL-S, FMQ Engine, FMQ Discoverer and FMQ Moderator that can be depicted as follow:

Figure 11: The Framework of FMG-QCMA

1. FMQ Distributor: enhances the capability of Quali-Fuzy Classifier in QCMA with following functions:

(1) All collected web service registered in the FMQ UDDI / OWL-S will be classified fuzzily according to fuzzy web service management performed in the FMQ Distributor. The rule of fuzzy classification on given web service will be analyzed by FMQ Engine.

FMQ Discoverer End User

End User Vague R equest QoS

Feedback

Vague R equest

QoS Feedback

Web Service s Discovered

Web Services Discovered

FMQ-Inference Rules

FMQ Engine

Web Service Information FMQ UDDI

OWL-S

FMQ Distributor

FMG OWL

Web Services R egistration

Web Services Information

Request / Reply Update Fuzzy Classification on Corresponded Web Services

Se manti c Analysis

Web Services Provision (for Classification)

R ules / QoS Analysis Rules

Analysis

Fuzzy Opi nions Requirement

FMQ Moderator (with FMGSAM)

FM QoS Administration

FM Consensus Analyzer

Rul es Analysi s

(2) Interpreting fuzzy web service inquiry issued from FMQ Engine, the FMQ Distributor reasons the fuzzy web services, retrieves the required web service stored in FMQ UDDI / OWL-S, returns the required web service back to FMQ Engine, and updates the correlated QoS status stored in FMQ UDDI / OWL-S in FMG-QCMA.

2. FMQ UDDI / OWL-S: is derived from UDDI / OWL-S in QCMA for registering and storing the web service which is provided from web service providers (vendors). Besides the web service registration, there are two major operations designed for fuzzy web service handling and corresponded classification:

(1) The fuzzy classification for registered web service, which is updated in FMQ UDDI / OWL-S, will be moderated by FMQ Distributor by event (driven by analysis from FMQ Engine).

(2) The definite web service exploration from service consumers will be performed via FMQ Discoverer. Any well defined requests from service consumers will be issued from FMQ Discoverer and being dispatched to FMQ UDDI / OWL-S directly rather than fuzzily analyzed through FMQ Engine, from viewpoint of FMQ UDDI / OWL-S.

3. FMQ Engine: extends the capability of the Quali-Fuzy Engine in QCMA with the following functions:

(1) FMQ Engine analyzes the vague inquiry or the fuzzy QoS opinions (when service consumers set his/her disposition on each QoS attribute and preference order over QoS attributes) received from the FMQ Discoverer and reasons over the vague inquiry using fuzzy logic. The rules to interpret the vague inquiry from FMQ Discoverer are stored in object FMQ-Inference Rules.

(2) FMQ Engine ascertains to which fuzzy QoS opinion sub-group the user making the inquiry to FMQ Discoverer belongs, QoS attribute reasoning in similarity and QoS

attributes preference order via invoking QoS analysis in FMQ Moderator.

(3) FMQ Engine asks for retrieving the fuzzily classified web service managed by FMQ Distributor by inquiry from FMQ Discoverer. The recommended web services for the inquiry will be returned to FMQ Discoverer after FMQ Distributor replies to FMQ Engine.

(4) FMQ Engine helps to fuzzily classify web service that FMQ Distributor gains from web service providers. The semantic analysis for the request of fuzzy classification will be performed via invoking the rules defined in FMG OWL.

4. FMQ Discoverer: is the object of Man-Machine interface which handles web service inquiries and fuzzy QoS opinions from service consumers and recommends right web services accordingly. The major operations designed in FMQ Discoverer including:

(1) FMQ Discoverer receiving all vague requests (fuzzy inquiry or setting of fuzzy QoS opinions) from service consumers for the selection of the appropriate / recommended web services, completely same as what Quali-Fuzy Discoverer did for QCMA.

Definitely, the vague requests will be converted as fuzzy requirement which will be delivered to FMQ Engine for further rule analysis. However, if the requests from service consumers are decoded as well defined requests rather than vague requests, then the “well defined requests” will be converted as a definite inquiry and delivered to FMQ UDDI / OWL-S for looking up the web service directly.

(2) When FMQ Discoverer receives vague request (including vague inquiries or fuzzy QoS opinions that could be issued by service consumers), it will also request the later feedback from the service consumers’ perceptions and opinions on QoS in order to modify service definition after locating and selecting the required web services. The steps involved not only analyzing the semantic definition of each vague request, but

also examining the meaning of the required quality attribute which is represented in the vague request.

5. FMQ Moderator: is to improve the capability of the Quali-Fuzy Moderator in QCMA, especially for multi-groups framework of QoS consensus analysis, by including the following functions:

(1) For the fuzzy QoS opinions FMQ Moderator moderates the perception derived from service consumers for the potentially recommended web services deployed in FMQ UDDI / OWLS.

(2) FMQ Moderator initializes the FMGSAM operations, including the clustering of all fuzzy QoS opinions. All fuzzy QoS opinions and their temporary analyzed matrixes are stored in FM Consensus Analyzer.

(3) Via FMGSAM operations defined in FMQ Moderator, the AM, AAD, RAD, CDC and group consensus in FMGSAM for each collected / converted fuzzy QoS opinion are obtained.

(4) Verify the later feedback from web service consumers if his / her delivered fuzzy QoS opinion was clustered in right sub-group or not. If the number of later feedbacks for the fuzzy QoS opinions clustering reaches the threshold m_threshold_distortion through algorithm Clustering_Verification, then the fuzzy QoS opinions clustering will be identified as “Mismatched Similarity” and the whole

SQ

WSA will be re-clustered via algorithm Fuzzy_Clustering under the condition.

(5) To perform RMGDP for each clustered sub-group. The result of analyzing preference order over all QoS attributes for each QoS opinion sub-group can be obtained. The outcome of RMGDP will be delivered to FMQ Engine for further update on

mechanism of generating recommended web services.

(6) The FM QoS Administration provides the set of QoS attribute definitions and initial value of system parameters. In order to reach the consensus on the definitions of QoS attributes for grouped fuzzy QoS opinion of each sub-group, the service consumers’

subjective opinions and preferences over QoS attributes have to be registered and stored by FM QoS Administration.