QoS-Aware Semantic Service Selection: An Optimization Problem

Abstract

In order to select the best suited service among a set of discovered services, with respect to QOS parameters, a user have to state his or her preferences, so services can be ranked according to these QOS parameters. Current Semantic Web Services ontologies do not support the definition of QOS-aware user preferences, though there are some proposals that extend those ontologies to allow selection based on those preferences. However, their selection algorithms are very coupled with user preferences descriptions, which are defined without semantics or at a different semantic level than service functionality. In this work, we present a service selection framework that transforms user preferences into an optimization problem where the best service is selected. This framework is based on an ontology that conceptualizes these user preferences. Thus, we use a very expressive solution decoupled with the concrete selection technique by using XSL transformations, while describing QOS-aware user preferences at the same semantic level of functional preferences.

Qualitative preference-based service selection for multiple agents

“…Combine preference statements in all levels; 5: Sub_aSet T [i] = NULL; 6: Add the combination into T [i] as an item; 7: Add T [i] into aSet T ; 8: end for 9: for each remaining attribute do 10: if attribute value has strict priority then 11: Order preference statements by priority; 12: end if 13: Add them into each Sub_aSet in T ; 14 Compute weighted distance using Eq. (2); 9: Order them according to weighted distance; 10: end for…”

One Size Does Not Fit All: A Group-Based Service Selection for Web-Based Business Processes

“…In this context, the number of service instances of different service providers can also be different, i.e., j is not equal for all service providers. As an example, two different service providers which offer, respectively, two and three service instances for the Property Valuation functionality can be denoted as S 1 j = {P ropertyV aluation 11 , P ropertyV aluation 12 } and S 2 j = {P ropertyV aluation 21 , P ropertyV aluation 22 , P ropertyV aluation 23 } As previously mentioned, our group-based selection process considers the selection of a Web service instance for a particular service functionality. Since we are assuming that all candidate Web service instances that provide a particular functionality have the same set of QoS properties therefore, the set of QoS properties associated with a particular service functionality Sf m is captured by the following set:…”

“…Our group-based approach shows an optimal service selection, utility maximization of multiple QoS criteria, and cost minimization, can be achieved when considering group-based service selection based on customer classes and multiple provider offerings of variable QoS and cost values. Some other approaches [11], [12], [13], [14] consider the classification of users according to their preference/profiles, context and usage pattern in their optimal service selection approaches. Our approach has essential distinctions; 1) the selection is customized at a fine-granular level where process instances are grouped based on similar QoS and cost constraints that are significant to the organization who own the BP, 2) It considers not only different user classes grouped by QoS criteria but also various service offerings offered by the same and different service providers, 3) It dynamically combines different service selections from different providers offerings to achieve maximum utility and minimum costs.…”

Collaborative Approaches to Complementing Qualitative Preferences of Agents for Effective Service Selection

“…On another hand, qualitative approaches may be more feasible, natural and general [6], [4]. Garcia et al [8] present a service selection framework that transforms qualitative preferences into an optimization problem. However, the problem of incomplete qualitative preferences also exists.…”

“…Find the closest cluster, put a i in it; 6 Calculate precise of each cluster as the sum of 7 mean square deviation; if precise > α then 8 Choose CP-net with smallest average distance 9 to other CP-nets for each cluster as new center; else 10 Exit while loop;…”