Modeling Metrics for Service Interpretation

Information plays a majo r ro le in various application do mains like library, financial, Health Care and so on. Information as a service in these domains is achieved by applying Service Oriented Approaches. Handling information about those services are important in discovering the appropriate services for exact matching of consumer requirements. So the available info rmation about these services needs to be organized in a better way for efficient access. Interpreting the appropriate service fro m the service registry needs complete information of the service. Researchers have discussed basic forms of representing informat ion about services through functional aspects that help in identifying the required web service. This information addressed does not fulfil the consumer requirements normally; hence an extended registry has to be provided with additional details of non-functional aspects in order to locate the exact service. The effect of these attributes on discovering a required service has to be measured. This paper focuses on formulat ing metrics for interpretation of services based on functional and non-functional aspects of a service. Fro m the literature we have identified features for interpretation. These features have been considered as a focal point and a metric suite is proposed to address those features. Based on these metrics, a measure for service interpretability is proposed. To verify the effectiveness of our proposed metrics, an experiment has been designed and carried out. The result of the proposed metrics shows the effectiveness and improvement of service discovery which g ives exact matches to consumer requirements.


Introduction
Applications in all fields are being developed as Service oriented applications and have acquired dominance among development styles. Services portray either single or mu ltip le functionalit ies. Addressing mult iple functionalities could be achieved through service composition. Serv ice composition comprises of identifying the required services and combin ing one or more services to obtain a co mposite service. In o rder to co mpose the exact service, service discovery plays a major ro le in identify ing that required service. Service discovery is concerned with identifying the appropriate services for fulfilling consumer requirement [1] [3]. Effective service discovery is achieved through better interpretation.
Interpretability deals with understanding of service with reference to functional and quality of service Meta data. Hence service Interpretation needs sufficient docu mentation and relevant Meta d ata wh ich are used to interp ret appropriate services. Functionalities rendered by a service are described through interface defin ition and details about syntax and semantics of services available in the service registry. Quality of service informat ion is required to enhance discovery to suit the consumer requirements. The Qos information dwells with Availab ility, Co mpliance, Response Time, Throughput, Latency and Doc. Significance of service interpretability can be obtained fro m [20] [23] [26] [46]. A need for measuring interpretability becomes vital.
Much of the research contribution is towards addressing the metrics for functional aspects which measures the interface and semantics of the web services. Other researchers have proposed measures for certain quality of service aspects like availability and response time. Hence the measures to corresponding interpretability are in primitive stage.
In this paper we are focusing on identifying the features for both functional and non-functional aspects of services interpretation. We have proposed metrics for the aspects identified and finally we have defined a new metric for service interpretation based on the proposed metrics. In order to study the proposed metrics, an experiment was designed and conducted. The rest of the paper is organized, section 2 gives review of measures contributed for service interpretation, and section 3 elaborates the proposed work. The experiment design was illustrated in section in 4. The experimentation was carried out and results are reported in section 5. The conclusion is presented in section 6.

Related Works
One or more services provide related or common functionalities. It's hard to find out the exact service. There arises a need to define the information relevant to service which leads to easy identificat ion of required services. Service interpretation supports in searching and identifying the required service and also the measures corresponding to this component plays major ro le in service discovery. Our study concentrates on service interpretation of discoverability. The review has been categorized into three parts. The init ial part address the aspects related to service interpretation and the second part focuses on the measures contributed by researchers that have some relevance to interpretability aspects. The final part of survey delivers the existing measures specific to interpretability aspects.
The aspects addressed by various researchers relate to interpretation of services are shown in table in 1. Functional attributes such as syntax and semantics of services are discussed by [45] [19]. The non-functional aspects addressed by contributors are price, availability, response time, and throughput, reliability and network distance [24] [25][34] [45]. Some of the authors focus on enhancement or enriching the service registry additional attributes for better discovery [17] [34]. The aspects specific to interpretation of services are addressed. It emphasis need for measures and metrics, in order to verify the attributes.
The literature presented in table 2 delivers the existing works pertaining to measures proposed by different authors, which have some relevance towards interpretation of services.  [21] Proposed to approach based on recommendation system to provide quality of service information for assessing the behavioural and threshold policies of web services.

QOS
Natallia Kokash [19] Comparative study for choosing the effective approach from the existing approaches for finding the lexical and structural matching of web services.

Syntax and Semantics of web services
Young Kon Lee [24] Presents a classification scheme for representing quality data in service registry Modification of quality data in future to increase the accuracy.
Ahmed and Bernhard [34] Presented the list of Quality of Services attributes of web services and discuss about the importance of QoS attributes in service selection from the service registry Accounting, response time and availability.  [49] Presented the mediation approach to automatically identify the most appropriate Semantic web services for a given request Used Similarity measures to obtain relevant semantic web services Minghui Wu et al., [47] Semantic web service discovery method used to sort the list of web services which are retrieved by using similarity queries Functional semantics and non-functional semantics and proposed measure for semantic similarity using functional semantics Benjamin [15] Used similarity measures to specific elements in a WSDL document for ranking the web services from the list of Web services

Similarity measures for functional features
Kee-leong Tan [18] Proposed the model for checking availability to determine the availability status of mobile web services Non-functional aspect -Availability metrics Bensheng Yun [48] Combined approach of behaviour matching with fuzzy similarity are used for service matching Measure proposed for service matching Yu-Huai et al., [44] Proposed a hybrid approach for automatic discovery of web services. The discovery based on textual and ontology information about web services Proposed metrics service similarity, operation similarity and similarity of input and output. Contributed a discovery algorithm for service matchmaking which uses syntactic and semantic searches in service registry for getting accurate results

Syntax and semantics metrics
Hong et al. [23] proposed the quality of Service Data measures for filtering the web services Proposed measures for completeness, Timeliness and interpretability Ehsan Emadzadeh et al., [46] Proposed schema matchers techniques based on semantics and Quality attributes Measures proposed for syntactic, semantic, (Correctness) and quality aspects (completeness) The details expressed in the table 2 convey that the existing measures are focused mainly on functional aspects of services. The functional aspects taken for measures are either primitive or not specified exactly. And also researchers have talked about quality attributes but the measures corresponding to the attributes are not addressed. They proposed measure for few attribute (e.g. availability)[kee-lee]. These shows there should be need for exact measures for service interpretation.
Final part of the Review p resented in table 3 g ives the existing measures proposed by various researchers that are more specific in finding out the appropriate services fro m the service registry.
The literature reveals that measures specific to functional aspects are addressed with semantic and syntax metrics. These metrics are focused much towards the technical data representation of services (deeper about the technical informat ion about services i.e. validating the syntactic and semantic representation of functional data) and does not provide support to interpretability measures of services [16][26] [46]. Similarly the existing QoS attributes measures are limited [20] [23]. So me of the author have proposed metrics for attributes like availability, response time, throughput and reliability (i.e. measures are proposed only for limited attributes). Hence there arises the need to measure other attributes also. From the study we have found that interpretability metrics of services are not addressed correctly. The measures corresponding to functional and quality of service attributes are in primitive or early stage needs more explo ration.

Proposed Work
Discoverability is the process of searching the individual service based on the service description and to invoke or interpret those services based on the purpose and its capabilit ies [2]. Here the definition o f d iscoverability indicates that the two components or items, discovery and interpretability are involved in the entire process of service discoverability [5][10] [23][25] [32]. The discovery deals with the searching or finding the service and interpretability deals with usage or invocation of those services. So d iscoverability has to address these two components to offer better discovery. To address discovery and interpretability components we need to identify the features supporting these two items. In this paper our focus is to propose measures for interpretability co mponent of discoverability.

Service Interpretability
Interpretability of services deals with clarity or communicat ion wh ich uses the functional and quality of service data for invocation. To invoke or use the services efficiently the functional and non-functional aspects i.e. quality of service data of each registered services has to be defined or represented clearly [11][12] [13]. Fro m the study we have found out the factors or features which listed below are essential for the invoking the services.

Functional Specification
Normally the functional data of service depicts the purpose and capabilit ies of the services in the service registry [14]. The two co mponents which are used to represent functional data are [22] [31] • Semantic Elements -The semantic elements are used to represent the purpose of the service (i.e. This defines the scope of the services) • Service Operation -The syntax or interface, wh ich depicts the operation or capabilities of services (i.e. it clearly represents what functionalities are offered by services)

Quality of Service Meta Data
The Quality of Service data is used for finding the suitable service fro m the group of services which meet out consumer requirements. The Quality of Service data used by consumer for evaluating and filtering relevant service from group of services because it gives the behavioral characters, Operational thresholds and policies of the each service in the Service Reg istry [2].
We have identified the various quality data which are used by consumer fo r assessing or filtering their service are listed below [

Functional Data Measures
• Check for Described Semantic Elements Checking for Described Semantic Elements (DSE) is measured by assessing the ratio of matching semantic elements to total matching and mismatching semantic elements of Serv ice. This metric check whether purpose or scope of the services are described properly or not.

Ratio of Described Semantic Elements (DSE) =
Here the value range of DSE is 0.. The functional data value measure is calculated by using the values of two metrics. FDV is co mputed as Functi onal Data Value (FDV) = Here W1 and W2 is the weight factor whose value is 0.5. We are giv ing the equal weights to both factors because the two factors are essential. Service operation is important co mponent to expose the functionalities of service. Semantic elements are not a mandatory but it's used to increase the usability of services.

Quality of Service Measures
The Qos attribute measures for each quality are described below, here we have found out the expected minimu m and maximu m values for each quality attribute. The minimu m value is calculated as rat io of min value of each Qo S attribute to maximu m value of each QoS attribute. The maximu m value fo r each data is obtained fro m max of value of each quality attribute to max range of each quality attribute.
The value of numerator and denominator are taken fro m the service registry. Expected min imu m metrics values are used only when the particular quality of service data is not available in the service registry. The value range for these metrics falls fro m 0 to 1. In case of response time and latency ratios only we use expected max value (response time and latency) remain ing ratio's we have used expected minimu m only.
• Ratio of Availability (Avail) Availability of services is measured by using this metric,

Msemanticelements
Msemanticlements Where, Desired (Avail) is expected availability of service, Agreed (Avail) is the availab ility offered by the service Max (Avail) is the maximu m availability value for service Here value range of Availability is fro m 0 to 1. Higher the value of this rat io indicates high availab ility of services.
• Ratio of Co mp liance (Co mp) Where, Agreed (comp) is the comp liance offered by the service Max (co mp) is the maximu m co mpliance value fo r service Here value range of Co mpliance is fro m 0 to 1. Higher the value of this ratio g ives high compliance of services.
• Ratio of Response time (rt) Where, Agreed (rt) is the number of seconds taken by service to respond request Max (rt) is the maximu m nu mber of seconds taken by service to respond request Here value range of Response time is fro m 0 to 1. Lo wer the value o f this rat io depicts better response fro m services. We are normalizing the value to 1 because all the rat ios are in max value except two.
• Ratio of Throughput (tp) Here value range of throughput is from 0 to 1. Higher the value of this ratio indicates the services can handle more number of user requests.
• Ratio of Latency Where, Agreed Delay is the number of second's delay of service to respond request Max. Delay is the maximu m number of second's delay of service to respond request Here value range of latency is fro m 0 to 1. Lower the value of this ratio indicates the services offer less delay in processing requests. We are normalizing the value of latency to 1 • Ratio of Doc Here value range of Doc is fro m 0 to 1. Higher the value of this ratio indicates the services offer more documents for better usage.
• Ratio of Reliable Messaging (RM) Here value range of Reliable message is fro m 0 to 1. Higher the value of this ratio indicates the services can handle mo re Error messages.
• Ratio of Best Pract ices (BP) Here value range of best practices is from 0 to 1. Higher the value of this rat io shows the services adopted good practices. The overall quality of Service data Measure (QDM) is co mputed as Where, Qi gives ratio of each quality data We have used eight qualities of Service data, the maximu m value of i is 8.

Interpretation Metrics
Finally, the interpretation of service (IoS) is co mputed by the values of Functional Data measure and quality of Service data measure (i.e. FDV and QDM ).
Here value range of IoS is fro m 0 to 1. Higher the value of this measure gives better invocation of Service.

Experiment Design
To demonstrate the usability of the proposed metrics, we have designed and imp lemented three different service registries. Each registry contains three different ranges of data (i.e. registry with 1000, 2000 and 3000 entries) [14] • SR1-Serv ice Registry 1 is the basic registry which contains limited nu mber of attributes • SR2 -Serv ice Registry 2 extended version wh ich contains additional attributes when compared to SR1.

Service Registry Attri butes
The attributes chosen are based on the review of various works and the values for each attributes are defined with help of the references and few attributes are defined by our self that are checked for its optimu m. The in formation given below gives description about each attributes and corresponding values for them.
• Registry attributes listed in table 4 describes the complete information of each reg istered service. Here the attributes are differentiated based on functional and quality of service data. The primit ive attribute is service name usually represented using the string type. Service category provides the support for better organization of services and to avoid the misplace of services falls under string type, service version is a nu mber type attribute allowing for simu ltaneous deployment of multip le versions of the same service and allowing the consumer to choose the version he wants to use. An interface is a fully qualified name of the service, ensuring that a consumer refers to the interface what the services actually expose.
The Consumer Type parameter allo ws us to assign different service endpoints/bindings to different types of consumers for example p latinu m/golden/etc. The other fields or attributes like semantic elements and service operation falls under type number and are used to represent the purpose and capabilities. The Semantic elements give the described semantic elements matching to consumer demands or requirements. The attribute value is set to max o f 5 and min of 3 for our experimental purpose. We have checked the optimality for these values. Service Operat ion gives the number of operation defined for the service. The attribute value is set to max of 6 and min of 3 for our experimental purpose.
• The quality of Service data list out the various fields and their values for the services in the reg istry to filter and use appropriate services that matches the service consumer demands. The values for each attributes and units are chosen based on the references [6][10] [11][24] [25].
The Service registry SR1 is designed with minimu m or basic fields and SR2 with additional fields other than SR1 and SR3 is the comp lete set which consists of the all fields defined in the table wh ich is exp lained separately in section 4.2, 4.3 and 4.4. Here we have considered the banking and financial services (B&F Services) as specific category for conducting the experiment towards interpretation.

Interpretation Metrics on Service Registry1 (SR1)
The Service Registry (SR1) contains limited attributes. It contains basic attributes like service name, category, service ID, service operation, availability and compliance. Here we formulated 12 queries for our experiment.
i.e. Query1 contains Category, Query2 contains Category + Co mpliance, and likewise remaining queries contains the fields fro m prev ious queries in addition to its own field.
Out of 12 queries, SR1 g ives response for first three queries and for the remaining queries values of query 3 will be repeating as it is a primit ive registry and contain basic fields.   The QDM is co mputed by using the two QoS attribute measures (i.e. availab ility and compliance) as shown in table 6. The remaining field measures values are computed by using the expected min imu m and expected maximu m met ric.
Here the expected minimu m is not applied Latency and Response time because for these measures expected maximu m is the worst case. For remain ing quality data the worst case is expected min imu m.

Interpretation Metrics on Service Registry2 (SR2)
The Service Registry 2 (SR2) is the extended version of SR1 with additional attributes like version, interface name, Response time and throughput. Out of 12 queries, SR2 g ives response up to the sixth query and the remaining there is no response, the values of the query 6 will be repeating because it is an extended version which contains additional fields compared to SR1. In case of SR2 the DSO met ric will be high when co mpared to SR1 because it has an additional attribute versioning of services. The versioning of services will have an impact on these defined service operation. Hence the FDV values of SR2 are high.

Interpretation Metrics on Service Registry3 (SR3)
The Service Registry 3 (SR3) contains the all attributes listed in the table 1 because it is a co mp lete reg istry and gives output for all the 12 queries. The FDV is co mputed based on two factors but in the case of SR1 & SR2 it uses only defined service operation (DSO). Similarly in the case of QoS data measure value is calcu lated by using the values of all quality of service data measures.

Findings & Discussion
The experiment was conducted against the three different registries that have been formed with B& F services, by using certain queries which supports interpretation. In analysis, we focus on each metric value that is applied in the experiment. The table 11 displays the result of FDM values of three different registries. High value FDM shows that the services contain more functional data i.e. the semantics and syntax of services are clearly defined. Consider the B&FI services 13 , the FDM value upon three different registries indicates there is a gradual increase because the clear representation of syntax and semantics of the service. In the case of B&F Serv ice 32 there is a sudden increase in FDM value on service registry3 because semantics are exp ressed more precisely when co mpared to other two registries. So the complete/essential information about syntax and semantics has greater importance in FDM value as shown in figure 1. This indicates that high value of FDM gives better interpretation of Serv ices. Table 12 depicts the results of the QDM values of three registries for various Serv ices. Here Qo S data measures values shows an impact of presence of more quality attributes (i.e. service registry contains more quality attributes acts as the filter prov ide effect ive interpretation). Here the services 13, 32, 91 g ives the gradual increase in the QDM value due presence of various additional QoS attributes in different service registries. QDM value for all services considered is high in case of service registry SR3 as shown in figure 2.  The effect of FDV and QDM values for measuring service interpretation upon various registries is depicted in table 13. It indicates that SR3 gives more Io S values when co mpared with other two registries. Figure 3 shows that Service Registry 3 B&F service 32, the IoS value is high when compared with other services but in remain ing registries there is a steady increase of IoS value among these three services. This sudden increase is due to inclusion of semantics value of the service. Fro m th is experiment we have observed that interpretation of services (IoS) is effective when a service represents its functional and quality aspects clearly and co mpletely. This in turn leads to better discovery of services.

Conclusions
We have designed the metric fo r Interpretation of Services (IoS) by proposed measures for functionality aspects and qualities aspects of Services. These measures are used in the experiment which was designed and conducted. The results are used to co mpute the IoS value o f services. Fro m the cases, it is evident that service reg istries contain essential informat ion about the service have higher impact on the IoS value. Higher IoS value indicates better interpretation of services. This has been experimentally proved fro m the values of metrics obtained for various service registries (SR1, SR2 & SR3). The Serv ice Registry 3 (SR3) gives better response towards interpretation of services as proved by the values of the metrics. Th is metric will help the Serv ice Provider to quantify the effect ive providing the essential informat ion about the services which will in turn enhances the discoverability of SOA systems.