WO2008111031A2

WO2008111031A2 - An improved grid computing architecture and method for invoking network services for subscription

Info

Publication number: WO2008111031A2
Application number: PCT/IB2008/052201
Authority: WO
Inventors: Sukesh Garg; Shishir Garg; Frederick Lee
Original assignee: France Telecom
Priority date: 2007-03-12
Filing date: 2008-03-12
Publication date: 2008-09-18
Also published as: WO2008111031A3

Abstract

A mediator host is provided for identifying services of networked resources which are available for invocation in a grid computing architecture. A mediator host Application Program Interface (API) is configured to receive a service registry query and related resource query. The API identifies the service registry query and related resource query by signature. A service registry query processing agent is configured receive an identified service registry query from the API, and, to communicate with a corresponding service registry to satisfy the query. A resource query processing agent is configured to receive an identified related resource query from the API, and, to communicate with a corresponding resource registry to satisfy the query. The mediator host returns a service query response which satisfies a predetermined resource metric to identify candidate services for invocation.

Description

WF, QFTHE INVENTDN

ANIVIHOVEDGBroCOMPUIINGAHDHnBCTLKEAND MEHOD R)RlNVOKING NEIWORK SERVICES roRSUBSCiaPIION

BFfATFD APPIIC ATIO NS

This application is related to co-owned and co-pending U.S. Patent Application SeiialNo.11/364,204 filed on March 1^st, 2006 entitled "A GRid computing architecture and associated method of invoking/Registering network services for subscription", the foregoing incorporated herein by reference in its e ntire ty.

BACKGROUND OFTHE INVENTDN

The present invention relates to a grid computing architecture and, more particularly, to a grid computing architecture and associated methodolgy for programmatic ally invoking/registering ne two rk se ivic es fo r sub sc rip tio n. The "background" description provided he re in is fo r the purpose of generally p re se nting the c o nte xt o f the inve ntio n. Wo rk o f the p re se ntly na me d inve nto rs, to the extent it is described in this background section, as well as aspects of the description which may not otherwise qualify as prior art at the time of filing, are neither expressly or imp lied Iy admitted aspriorart against the present invention. An executable service is a set of related executable functions which can be discovered orcalled (ie., "pro grammatically invoked") via an established network protocol. Such services include World Wide Web based services which are increasingly employed in carrying out the func tio ns o f c o mp o site applications. The leveraging of network distributed web services to function together as a composite application isreferred to asgrid computing ordistributed object computing. h operation, a grid computing architecture employs the web services as an application integration technology. Fa ch web service is supported by a resource of the grid computing network. For example, in an Internet grid computing environment, individual web services may be supported by an execution environment, such as a corresponding web server, which is the resource for the server. The information necessary to programmatic ally invoke the web service is defined in a Web Services Description Language (WSDL) document or Interface Definition Language (IDL). These documents are stored in a registry of the grid computing framework so that applications seeking subscription to a specific service can locate and invoke the necessary service. One wellknown type registry of web services is the Universal Description, Discovery and Integration (UDDl) registry. This registry type would include the location of the service and the necessary information for integrating this service in an application of the grid computing framework. Using the necessary information, an application can remotely call the service though an Extensive Mark-up Language (XML) based messaging protocol such as the Simple Object Ac ess Protocol (SOAP) or Object Request Broker (ORB) of the Common Object Re quest Broker Architecture (CORBA).

Efficient discovery of grid services and resources is an essential build ing block fora grid computing framework. The prevalent standard forweb service discovery is UDDL In order to gauge the resource supporting the web service, typically, the GIo b us To o Mt s Mo nito ring and Disc o ve ry Se rvic e (MDS) is utilize d . The c o rre sp o nding G Io bus To o Mt' s MDS registry is pa rticulariy suited for registering resources.

Yet, grid computing environments typically offer numerous resources with no indie ation of the service availability or re source information where these services are available. It is practically impossible to choose a resource for subscription because there is no way to determine whether the resource is be ing used by o the r se rvic e s or whether the resource is heavily utilized or underutilized. Rirthermore, whenever a user or an application wants to query a service and find available resource for subscription, service discovery has to be performed first, and thenMDSmaybe used to query foravailable resources. h US Patent 11/364,204, the applicant presented an API (Application Program Interface) Mediator service that offers a single API for registering and discovering both resourc es and servic es. The APIis robust enough to allow fora unified query for b o th se rvic e s and re so urc e s o n whic h the se rvic e s are running . It allo ws fo r que rying a resource with a given resource property in a specific range and at the same time ensures that the requested service is available. This converged approach considerably reduces the result set by eliminating services which are out of scope or resources that are not available. An example of such a query is "End a Translation Service on a Resource where C PU utilization is less then 40%". The mediator service leverages the UDDI registry and the MDSregistry to fulfill the request.1 uses the UDDI registry for service discovery and the MDS directory for resource discovery. The unified API leverages the best in class APIs (UDDI and MDS) while hiding the complexity from the user. However, the API mediator disclosed in US ll/364,204only permits the capturing of syntax information associated with b o th the service and the re so urc e .

Some services may be published with the same syntax, while they have different profiles and capabilities. The API mediator here above would return all these services even though some of them mightnotbe relevant.

Presently, there is still a need for an interface service that transparently mediates queries fo r id e ntifying and/ or registering both resources as we 11 as services associated with the re so urc e s, while taking into account the profOesand capabilities of both the services and resources that are queried. Such an interface would only return services and resources that are really relevant to the requester of such queries.

SUMMARY O F THE INVENTD N

The present invention provides a mediator host provided for identifying services of networked resources which are available for invocation in a grid computing architecture. A mediator host Application Program Interface (API) is configured to receive a service registry query and related resource query, said queries comprising profile information about the services and related resources to identify, said API id e ntifying the service registry query and related resource query by signature. A service registry semantic querying agent is configured to receive an identified service registry query from the API, and, to performa semantic analysis in a corresponding service registry to satisfy the service registry query, said semantic analysis being performed using the profile information for the services to identify. A resource querying semantic agent is configured to receive an identified related resource query from the API, and, to perform an semantic analysis in a corresponding resource registry to satisfy the related resource query, said semantic analysisbeing performed using the profile information for the resourcesto identify. ha further embodiment of the mediatorhost according to the inve ntio n, the mediator host returns a service query response which satisfies a predetermined resource metric to identify candidate services for invocation. ha furtheraspect of the invention, a grid computing architecture isprovided forinvoking services of networked resourcesto execute a composite application. A client node of the architecture isconfigured to provide a service registry query, and, an associated service resource query of the grid computing architecture, said queries comprising profile information about the services and related resources to invoke. The service registry queiy and associated resource queiy may have different message signatures. A mediation node includes an application program interface (APD which is configured to receive the service registry query and associated resource query. The API identifies the service registry query and associated resource query by signature. A service registry semantic querying agent is configured to receive an identified service registry query from the API, and, to perform a semantic analysis in a corresponding service registry to satisfy the service registry query, said semantic analysis being performed using the profile information for the services to invoke. A resource semantic querying agent is configured to receive an identified associated resource query from the API, and, to perform a semantic analysis in a corresponding resource registry to satisfy the related resource query, said semantic analysisbeing performed using the profile information for the resourcesto invoke. h a further embodiment of the architecture according to the invention, the mediation node returns a service query response which satisfies a predetermined resource metric to identify candidate services for invocation.

Instilla furthe r a sp e c t of the invention, a computer readable carrier includes computer program instructions that cause a computer to implement a method of identifying services of networked resources available for invocation in a distributed object computing architecture. A service registry query and related resource query are received at an API, the service registry query and related resource query have different signatures. The service registry query and related resource query are identified by signature. The identified service registry query of the APIis processed by performing a semantic analysis in a corresponding service registry, said semantic analysisbeing performed using the profile information for the servicesto identify. The identified related resource registry query of the APIis processed by performing a semantic analysis in a corresponding resource registry, said semantic analysisbeing performed using the profile information for the resourcesto identify.

1 is to be understood that both the foregoing general description of the invention and the following detailed description are exemplary, but are not re stric tive , o f the inve ntio n.

BRIEF DESCRIPTION O F THE SEVERAL VIEWS OFTHE DRAWINGS

A more complete appreciation of the invention and many of the attendant advantages thereof will be readily obtained as the same becomes better understood by reference to the following detailed description when considered in c o nne c tio n with the accompanying drawings, wherein:

Figure 1 is a high level block diagram of a computing architecture in accordance with an exemplary embodiment of the invention; Figure 2 is a more detailed block diagram of the mediator interface of the exemplary embodiment of Figure 1;

Figure 3 is a diagram describing the resource ontology implemented in the media tor interface of the examplary embodiment of Figure 1;

Figure 4isa process flow ofa new service and related resource registration process in accordance with an exemplary embodiment of the invention;

Figure 5 isa process flow ofa service query in accordance with an exemplary embodiment of the invention; and

Figure 6 is a plot showing hit ratios for the same service and resource query using different media tor APk

Certain terminology used in the following description is for convenience only and is not limiting. The term "service" as used herein, is not limited exclusively to a single function, but embraces sets of related functionality. likewise, the term "resource" as used he re in is no t limite d to a single metric ofa related service, but, instead, also embraces a set of metrics which describe the operating environment ofthe related service. In the drawings, the same reference numerals are used for designating the same elements throughout the several figures.

The present invention provides efficient disco very/ registration of grid services and associated resources, h a grid computing architecture or distributed object computing environment, a client device or application invokes web services for executing a composite task Web service communities have addressed the need forservice discovery, via an industry standard called universal description, discovery integration (UDDl). UDDI registries can be either public, or private. Fbr example, enterprises may deploy private UDDI registries for use within their organizations, h addition to the registration of services, there are registries which include information on service resources, such as the web services version of the Monitoring and Discovery Service (WSMDS), also known as the MDS4 component ofthe Globus Tb o Mt version 4 (GT4). Service Discovery isa critical component required to d e live r o n the promise of truly loosely coupled Service Oriented Architecture where applications can dynamically discover and use available services on the web. However, typically these so it o f se rvic e discoveries need to not only identify available services from a feature and capability perspective but also the underlying performance and availability of the re so urc e s o n whic h the se servicesare running, therefore impacting the overall complete performance of the service and resulting composite application.

Because applications and resources are distributed globally and across different virtual organizations that are inherently heterogeneous, Service Discovery becomes extremely important. The current approaches to service /re source discovery such as UDDI or MDS do not have for that matter sufficient expressiveness and efficient matchmaking abilities in their query language. A relational query language would enable more expressiveness but would be difficult to implement in scale. h order to determine resources and services, users need to not only use resource functionality information in finding the right service but also need to apply their own selection policies with regards to non-functional characteristics of services such as reliability, invocation cost, provenance, quality of service, reputation, etc . A service discovery language needs to be expressive enough to capture such metadata. Furthermore, this needs to be amenable to automatic processing. I Architecture

Referring now more specifically to Figure 1, a high level block diagram of a grid computing architecture 5 in accordance with an exemplary embodiment of the invention is shown. The architecture includes a service registry 7, a mediator interface 10 (Le., information mediator) and a resource registry 12.

The exemplary architecture 5 converges both service and resource registries, with both solutions working in a symbiotic relationship. Services are registered in service registry 7 and resources are registered in resource registry 12. The architecture 5 employs the mediator interface 10 to present services and resources in a distributed system using a unified application program interface (APD. Both service and resource registry comprise pro file information for registered services and resources, ie. information describing their capabilities as described here after and that c an be queried by semantic querying agents. Thanks to semantic querying agents, one for service querying and one for resource querying, the architecture 5 according to the invention can perform semantic querying of both service and resource registries.

As outlined below, UDDl, a platform independent XML (Extensible Markup Language), is the standard to register and query services stored in a registry. Exemplary UDDI registry has a well structured information model that effectively represents the relationship between the business, service and technical information. UDDI registry does not effectively represent dynamic state information for services and resources. A UDDI registry enables a registrant to identify types of information therein - white page data, yellow page data and green page data, h the yellow pages, registrants register an identity and their service sunder different categories, the same as in the fa miliar tele phone yellow pages, h UDDI here too, the white pages are a listing of entities. The green pages represent the technical information that is necessary to invoke a given service. Thus, by browsing UDDI registry, an application developer should be able to locate a service and a company and find out how to invoke the service. Furthermore, UDDI has been designed to support not just developers or users but also client applications that can dynamically, in real time, searchforand consume services pro grammatically with no manual intervention. UDDI offers APIs to publish and inquire with respect to web services. These

APfc, orquerying agents, are quite simple to use as the UDDI registry itself is exposed as a web service using the Simple Object Application protocol (SOAP). Of course, those skilled in the art will recognize that other registries may employ alternative XML envelopes such as DME; W3C and CORBA. However, the application of UDDI is not well suited for registering and management of state ful resources as UDDI does not represent service capabilities Therefore it is of no use forlocating se rvic e s o n the basis of what they provide.1 is ne ithe r suite d in its present state to register profile information for services.

The CMU (from the Carnegie Mellon University) Semantic Matchmaker is an entity that allows web services to locate o the r se rvic e s, provides a solution to the problem of matching services, and allows for full implementation of inte rope rate service providers on the web. It introduces OWL-S (OWLService), a OWL(Ontology Web Language) based language for describing service capabilities or profiles. Thanks to a semantic matching between advertisements and requested services, web services are located on the basis of the capabilities they can provide. The solutionto this matching usesa language to express the service capabilities, and the specification of a matching algorithm between service advertisements and service requests, one that re cognizes when a request matches an advertisement. OWLS has been adopted as a service description language, as it provides a semantic ally -based view of web services, including the abstract description of the capabilities of the service, the specification of the service interaction protocol, and the actualmessages thatit exchanges with otherweb services.

Standards such as SOAP and WSDL(Web Services Description Language) are designed to provide descriptions of message transport mechanisms, and for describing the interface used by each service. However, neither SO AP nor WSDL a re of any use forproviding the automatic location of web se rvic e s o n the basis of their capabilities.

Through the tight connection with OWIJ-OIL (Ontology Interchange Language), OWLS supports the need forsemantic representation of services. The main limitation of OWLfOIL is its lack of a definition of rules and an associated reasoner. Therefore, OWLS was coupled with RuIeML RuIeML can describe constraints related to input and output, and also preconditions and effects for planning. The CMU Semantic Matchmaker (also called here after the OWLS

Matchmaker) is also a web service that helps make connections between service requesters and service providers. The Matchmaker serves as a "yellow pages" of service capabilities. The Matchmaker allows service requesters such as users and/or software agents (ie. applications) to place queries for services and find matching services by providing a mechanism for registering service capabilities. Registration information is stored as advertisements. When the Matchmaker agent receives a query from a user or another software agent, it searches its dynamic database of advertisements for services that can fulfill the incoming request(s). Thus, the Matchmaker also serves as a liaison between a service requester and a service provider.

The CMU Semantic Matchmaker, acting as a service registry semantic querying agent, employs techniques from information retrieval, Aitific ial Site Dig e nee, and software engineering to compute the syntactical and semantic similarity among service capability descriptions in a corresponding service registry such as service registry 7. The matching engine of the matchmaking system contains different filte is for namespace comparison, wo id frequency comparison, ontology similarity matching, ontology subsumption matching, and constraint matching. The user configures these filters to achieve the desired tradeoff between performance and matching quality.

Resource registry 12 provides a resource registration and discovery service. A function of registry 12 is to provide a mechanism to query and update resource information. Exemplary registry 12 is an ideal way to represent dynamic attribute of resource, however, this information mo del is too simplified to show a variable aspect relationship. As seen here after, the Globus MDS registry may be used as to implement re source registry 12.

The Monitoring and Discovery Service (MDS) from Globus Toolkit is a set of web services for monitoring and discovering resources in a grid computing architecture 5. A query typically returns a set of resources that can complete a job in shortest amount of time with the most efficient use of resources. The resource selection process involves using static information such as CPUs, clock speed, physical memory, virtual memory and diskspace. In addition, dynamic information is also used such as current available CPUs, numberof jobs queued and current CPU and memory utilization. The MDS service forgathering the resource information is the Index Service. It collects the information and publishes it as a resource property and provides it via a web service. At any given time, the information presents the current configuration and state of the grid resources. I is maintained as a set of Web Service Resource Framework (WSRF) resource properties. MDS also provides subscription and notification on a resource property using the TriggerService. It can execute policies, once a specified threshold has been exceeded. The Index service and the Trigger services are built on the MDS Aggregation Framework. The framework offers a hierarchical index service registration enabling the consolidation of data from multiple locations into a single point. Queries can be made against the resource attributes which re present the state and configuration of the resource.

The MDSresource attributes are maintained in a "soft state" by associating a lifetime to it. This ensures the removal of stale infoimation. Since MDS maintains configurations and services in a soft state, constant updates are required in oiderto renew/update these services. Services are registered using the service registration file. The service registration file contains the grid resource, the service group the resource should register with and service configuration parameters.

MDS does solve the resource discovery and monitoring problem. However, MDS does not offer the flexibility and functionality offered by UDDI for service discovery. Fbr example, there is no taxonomy available in MDS such as UDDIs "category bag". Furthermore, there is no simple way to make a query for a specific service where the resource attribute falls in a specified range. MDSneeds complex xPath query language in order to identify a specific service.1 is not sufficient to simply identify a resource offering the service witho ut id e ntifying the resource that is capable of servicing the request. Fbr example, the existence of the service is irrelevant if the underlying resource ofthe service is incapable of servicing a request. Resource parameters which indicate such ability are dynamic in nature, for example CPU utilization or available memory.

An important aspect of the API, ie. the interface mediator 10 according to the invention, is the ability to semantic ally match the profile information for the services and resources to identify to the registered services and resources respectively. For that matter, a resource querying semantic agent is required, as the service registry querying agent is already provided through the OWL-S Matchmaker agent. The known Globus Toolkit MDS pro vide s a resource registration and discovery service. The main function of MDS is to provide a mechanism to query and update resource information. MDS is an ideal way to represent dynamic attributes of a resource (CPU, ...) but its information modelin its present state is too simple and too limited to introduce a profile for the resource. Furthermore, the current implementation of CMU Mate hMaker engine does not contain an ontology for resources since it only provides semantic service discovery, h orderto develop the unified semantic discovery interface according to the invention, a resource ontology is provided so that the queries for services and related resources comprise profile information formulated in a unified manner, e.g. using a unique OWL-S query.

There is an on going effort to define metadata information model on distributed computing system resources called Common Information Model (CM) lead by the Distributed Management Task Fb re e (DMTF). The Common Information Model (ClM) is conceptual info imation model for describing computing and business entities in internet, enterprise and service provider environments. It provides a consistent definition and structure of data , using object oriented techniques. The CM includes expressions for common elements that must be clearly presented to management applications like object classes, properties, methods and associations to name a few. CM uses a set of terminology specific to the model and the principles of object oriented programming.

The Common Information Model (CM) provides a method for describing and interacting with many different kinds of resources in an heterogeneous IT environment. Using CM, systems management applications have a single modelfor communicating with these different IT re sources. In the architecture according to the invention, CM model is used to build the ontology for resources as described here after. DMTF is associated with GGF (Global Grid Fo mm) to introduce CM into GRID

Open Grid Services Architecture (OGSA) scheme. In the mediator according to the invention, CM is used to describe ontology for GRID resource as seen in FJGr.3. CM provides a common definition of management information for systems, networks, applications and services, and allows for vendor's extensions. CM common definitions enable to easily define semantically rich metadata information for the GRID resources.

FIG.3 is an exemplary embodiment of a CM representation of a resourc e. The resource information may be organized in a tree line manner with the core element C M_C ORE corresponding to a given resource organized in two branches. A device branch CM_device is used to describe the logicaldevice of the resource: the type of processor, its clockspeed, its load percentage, and the type of storage device, diskdrive, maximal size, percentage of used disk, ... The second branch CM_system describes the system itself, type of operating system (OS), version of the OS, ... Any additional characteristic and properties may be added to further describe the profile information of the registered resources. CM_CORE, CM_device, CM_system and the likes may be seen as CM classes.

The resource ontology contains the concepts and characteristics of the resource as specified by the CM classes. And these characterize the resource for further advertisement, discovery and matchmaking. Mediator interface 10 can access the appropriate resource properties in each class by following logical ontology tree of the CM representation.

The resource ontology proposed here is compliant with the OWL-S language. Thus, by using the resource profile asde scribed here above, a client may compose a single querying ordiscovery message which is compliant with the OWL-Slanguage and then complete the entire discovery within a single transaction with the interface 10. The introduction of ClM model allows to augment the MDS registry with profile information, ie. capability description for registered resources, and perform a semantic matchmaking between the resource requester and resource provider, much alike the CMUsemantic Matchmakerforregistered services.

One p o ssib ility to built a resource querying semantic agent is to built from the known Globus Toolkit MDS agent a MDS matchmaker similar to the OWL-S Matchmaker; Such MDS matchmaker would provide a semantic querying of the augmented MDSregistry based on the profile information for searched resources h the mediator interface 10 according to the invention, the OWLS

Matchmaker is associated to the resource querying semantic agent through said interface 10 in a symbiotic relationship. Services are registered in the OWLS matchmaker registry 7 and resources are registered in the augmented MDS (using CMI classes and properties) registry 12 as seen in EIG. 1. The architecture 5 according to the inve ntio n use s the mediator interface 10 to present services and resources in a distributed system using a unified semantic API leveraging OWLS to provide service information and MDS to provide resource information and mediating as described herein improves service /re source discovery and registration. h the media tor interface according to the inve ntio n, the architecture 5 builds an overlay network of OWLS and augmented MDS registries based on the mediator interface 10, where the interface 10 acts as a rendezvous networkthat connects the multiple registries. The interface 10 sits at the very front of the architecture 5 to process all of the incoming messages as well as redirection to the appropriate querying sub -func tio ns. As outlined in more detail below, those skilled in the art will recognize that alternative registry types are embraced by the exemplary architecture 5.

The exemplary mediator interface 10 provides a unified and e fϊϊc ie nt unifo rm distributed service API The mediator interface 10 leverages the fact that every service has an end point identified by the physical location of the service. The end point is modeled as a resource, leading to the concept that eveiy service is associated with a resource. Furthermore, a resource can exist without a service. Referring now to FIg.2, a more detailed diagram of mediator interface 10 is shown embodied in a grid service container framework 9. The interface 10 mediates the communications between client 15 and backend registries such as OWLS Matchmaker registry 7 and augmented MDS registry 12. The interface 10 communicates with the OWL-S Matchmaker and augmented MDS registries using different pro to c o Is. As such, incoming messages 14 to the mediator interface 10 are consumed by a specific registry agent. h the exemplary embodiment, two semantic agents are provided, namely a "Matchmaker Mediator" 18 ( or UDDI mediator) and "Resource Mediator" 20 (or MDS mediator). As the name refers, Matchmaker Mediator 18 processes OWLS related messages and Resource Mediator 20 processes resource related messages. Both mediators are semantic mediators, Le. they can perform semantic querying in the respective registries. This pluggable framework guarantees maximum flexibility to enable the addition of an auxiliary specification to the existing mediator framework without any major modification. Fbr example, XXX mediator 22 and registry 24 illustrated in phantom reinforce the plugablity of the architecture 5. h the exemplary implementation, those skilled in the art will recognize that the architectural strength of this architecture 5 is the ability to add any kind of protocolto the framework with no changes required to existing clients, services and registries. With the abstraction layer formed by the mediator interface 10 upon heterogeneous registry environments (for services and resources), the client 15 is agnostic to whether the querying entry is for the Matchmaker registry 7 or the augmented MDSregistry 12. Using the exemplary embodiment of the architecture according to the invention, the message may be a OWLS based message, incorporating profile information for resources and/or services. The mediator interface 10 will transform any messages 14 passing there through to the appropriate protocol understood by the respective registries 7 and 12. The interface 10 is provided to the client 15 as a multiple registry implementation. No matter what registry is supported, client 15 can leverage the single API of the exemplary architecture 5.

Si the exemplary implementation according to the invention, the Matchmaker mediator 18, acting as a service registry semantic querying agent, is configured to receive an identified service registry query from interface 10, and, to perform a semantic analysis in service registry 7 to satisfy the service registry query. The semantic analysis is performed using the profile information of the services to id e ntify.

Furthermore, the resource mediator 20, acting as a resource querying semantic agent, is configured to receive an identified related resource query from the API, and, to perform an semantic analysis in a corresponding resource registry to satisfy the related resource query, said semantic analysis being performed using the profile information for the re sources to identify

An exemplary way to achieve the resource querying semantic agent 20 is to leverage the semantic capabilities of the OWL-S language. The information mediator 10, after receiving a OWLS message forquerying services and resources, uses the specific registry agent to consume an incoming OWLS message and sort out the service query from the resource one. The service query is forwa role d to the Matchmaker mediator for further processing. The resource query, whether at the information mediator or at the Resource Mediator level, is transformed into a WSRF (Web Service Resource Framework) message comprehensible by the Globus Toolkit. A resource semantic interpreter, or semantic parser, may be used to define resources with characteristics and properties semantically similar to the queries resources. Natural language techniques may be used to built a resource semantic interpreter, relying uponresource terminologies and characteristics.

From the searched resource profile information (in the initial OWLS message), as well as semantically similar profiles (provided by the resource semantic interpreter), a WSRF query string maybe built. The Globus Toolkit may then process the resulting WSRP query string by searching through the augmented MDS registry the resources matching the whole WSRF query. As a result, a larger number of resources may be retrieved through the Resource Mediator 20, as opposed to known MDS mediators. Indeed, with the use of the CMI ontologies for resources, which brings an enriched description of resources, a unique OWLS message, including a OWLS message for resources, can be formulated by a user. The mediator interface 10, after sorting out the service profile querying from the resource profile querying in message 14, uses the semantic capacities of both mediators 18 and 20 to retrieve mate hing se rvic e s and re so urc e s. h order to evaluate the performance of mediator interface 10, it is very important to capture the distinct characteristic of the dynamic resource. Considering effective query hit ratio can clearly demonstrate the efficiency of the me dia tor ho st ac cording to the invention. The effective query hit means finding real services with available re sources that the usermay invoke forgrid computing.

FIG.6 presents the query hit rate from a service discovery request using the proposed mediator host according to the invention, as well as known hosts. For the service query, it is assumed that: -all the services have diffe ient identity and equally distributed distinct profiles (e .g. a numberof 3, A, Band C, forthe results of FIG.6),

- each node (ie. resource) is randomly hosting one of the chosen services, -the service requestor (ie. the user or application calling the mediator host) wants to find a service profile (say A) which has C PU utilization of less than 30%,

- servic es are randomly distributed on the nodes, and,

- so me re so urc e s d o no t ho st any se rvic e .

As can be seen in FIG.6, the solid line (named UDDI query) shows linear yield of search result forthe service A using only a UDDI agent. FVe n though services have different capabilities, the UDDI query simply yields all the services cause of their identical service syntax. The square dot line (named MDS query) using the MDS toolkit agent shows almost linear yield of search result for all nodes which may or may not host the service, as no service query is taken into account. Only CPU utilization is handled. The dashed double dot line (name OWL-S query) shows reduced number of search result for the given service using only Semantic Mate hMaker agent. The OWLS query is able to narrow down the re suits be cause not all services have identical capability (ie. profiles). Nevertheless, no profile for resource is taken into account. The dashed dot line (named OWLS⁷MDS query) for the media tor host according to the invention yields the lowest hit ratio, as the search re suits take into account both the service and the resource profiles as defined by the user. The result is remarkable because it even reduced OWLS query result further afterapplying resource capabilities.

Thanks to the proposed architecture according to the invention, an effective narrowing down of the query results is achieved, which considerably increase the degree of relevancy of service discovery results. E Methodology

Any process descriptions in the process flow diagrams should be understood as representing specific aspects of circuit operation, or steps in the process. Alternate implementations are included within the scope of the exemplary embodiment of the present invention in which functions may be executed out of order from that shown or discussed, including substantially concurrently, depending upon the functionality involved, as would be understood by those skilled in the art.

As outlined below, process flow diagrams in accordance with an exemplary embodiment of the present invention are described relative to FKS.4 and 5 - adding a new service to an existing resource, and querying for service with resource properties respectively. Other operations like adding a new resource, querying a resource, updating, deleting a service /re source, etc. are similarly processed and are omitted forthe sake ofbrevity. The process flow of FIG.4 details a client 15 adding the service with an existing resource in the MDS registry 12. The service registration request leverages the existing resource entry and dynamically establishes the relationship between the service and the resource.

The client 15 sends service registration request to the information mediator 10 with existing resource information specified in the endpoint forthe service. Then, information mediator 10 analyzes the message signature and retrieves resource information (Le., which client to use.) The information mediator 10 forwards service registration request to Matchmaker registry 7 and waits for completion. Upon receiving the registration response from Matchmaker registry 7, information mediator 10 creates an association between new service and existing resource by sending new service key and service name to the resource.

The process flow of FIG.5 details a client 15 can query the service with a given set of resource properties. FJvery service has at least one resource associated with it. The operation returns the set of services that satisfy the query parameters which exist on resources which satisfy the resource query parameters. This operation is transparent to the client 15. The client 15 sends the service query to the information mediator 10. The information mediator 10 analyzes the request by looking at the message signature and determining the kind of request. Since the request is a service query it then, forwards the requests to the Matchmaker registries 7. On receipt of the response, the information mediator 10 forwards the resource query— with the requested resource properties to MDS registry 12. The result of resource query shows the set of resources which serve specific service and hold properties client is seeking. The information mediator 10 selects only services served by those re so urc e s fro m the service query result. And finally, returns the service list to the client 15.

The information mediator 10 can narrow down the service list b y c o mb ining resource information. That means client 15 can get the service list what it actually wants to use. h this sequence of operations, the client 15 was unaware of the association between the service and the resource and was able to get the info mi a tion request via the unified APIo f the info miation mediator 10.

Obviously, readily discernible modifications and variations of the present invention are possible in light of the above teachings. It is therefore to be understood that within the scope of the appended claims, the invention may be practiced otherwise than as specifically described herein. Fbr example, while described in terms of ha id ware /software components interactively cooperating, it is contemplated that the system described herein may be practiced entirely in software. The software maybe embodied in a earner such as magnetic oroptical disks, ora radio frequency oraudio frequency carrierwave.

Thus, the foregoing discussion discloses and describes merely exemplary embodiments of the present invention. Aswillbe understood by those skilled in the art, the present invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof. Accordingly, the disclosure of the present invention is intended to be illustrative, but not limiting of the scope of the invention, as we 11 as other claims. The disclosure, including any readily discernible variants of the teachings herein, define, in part, the scope of the foregoing claim terminology such that no inventive subject matter is dedicated to the public.

Claims

WHATE C LAMED E:

1. A mediator host identifying se ivic e s o f ne two ike d resources available for invocation in a grid computing architecture, comprising: a mediator host Application Program Interface (APD configured to receive a service registry query and related resource query, said queries comprising profile information about the services and related resourcesto identify, a service registry semantic querying agent configured to receive an identified service registry query from the API, and, to performa semantic analysis in a corresponding service registry to satisfy the service registry query, said semantic analysisbeing performed using the profile information for the servicesto identify; a resource querying semantic agent configured to receive an identified related resource query from the API, and, to perform an semantic analysis in a corresponding resource registry to satisfy the related resource query, said semantic analysisbeing performed using the profile information for the resourcesto id e ntify.

2. The mediator host of claim 1, wherein the mediator host returns a service query response which satisfies a predetermined resource metric to identify services available for invocation.

3. The mediator host of one of the previous claims, wherein the service registry query and related resource query is an Ontology Web Language Service (OWLrS) message.

4. The mediator host of one of the previous claims, further comprising a registry agent adapted to sort the service query fromthe resource query.

5. The mediator host of one of the previous claims, further comprising a semantic resource parser for interpreting the profile of the queries re sources from the resource query and building a string of Web Service Resource Framework (WSRF) queries comprising said queried resources and resources with profile semantic ally similar.

6. A grid computing architecture for invoking services of networked iesouicesto execute a composite application, comprising: a client node configured to provide a service registry query, and, an associated service resource query of the grid computing architecture, said queries comprising profile information about the service sand related re sources to invoke, a mediation node having an application program interface (API) configured to receive the service registry query and associated resource query; a service registry semantic querying agent configured to receive an identified service registry query from the API, and, to perform a semantic analysis in a corresponding service registry to satisfy the service registry query, said semantic analysisbeing performed using the profile information for the servicesto invoke; and, a resource semantic querying agent configured to receive an identified associated resource query from the API, and, to perform a semantic analysis in a corresponding resource registry to satisfy the related resource query, said semantic analysisbeing performed using the profile information for the resourcesto invoke.

7. The architecture of claim 6 wherein the mediation node returns a service query response which satisfies a predetermined resource metric to identify candidate services forinvo cation.

8. The architecture of one of the claims 6 and 7, wherein the service registry query and related resource query is an Ontology Web Language Service (OWLrS) message.

9. The architecture of one of the claims 6 to 8, further comprising a registry agent adapted to so it the service query fromthe resource query.

10. The architecture of one of the claims 6 to 9, further comprising a semantic resource parser for interpreting the profile of the queriesresourcesfrom the resource query and building a string of Web Service Resource Framework (WSRF) queries comprising said queried resources and resources with profile semantic ally similar.

11. A method of identifying se rvic e s o f ne two rke d resources available for invocation in a distributed object computing architecture, said method comprising the acts of: receiving a service registry query and related resource query at an API, said queries comprising profile information about the services and related resources to identify; processing the identified service registry query of the API by performing a semantic analysis in a corresponding service registry, said semantic analysis being performed using the profile information for the services to identify; and, processing the identified related resource registry query of the API by performing a semantic analysis in a corresponding resource registry, said semantic analysisbeing performed using the profile information for the resourcesto identify.

12. The method of claim 11, further comprising the act of returning a service query response which satisfies a predetermined resource metric to identify candidate se rvic esforinvo cation.

13. The method of claim 12, wherein the service registry query and related resource query is an Ontology Web Language Service (OWLS) message.

14. The method of one of the previous claims 11 to 13, wherein the service registry query and related resource query is an Ontology Web Language Service (OWLrS) message.

15. The mediator host of one of the previous claims 11 to 14, further comprising a semantic resource parser for interpreting the profile of the queries resources from the resource query and building a string of Web Service Resource Flame work (WSRF) queries comprising said queried resources and resources with profile semantic ally similar.

16. A computer readable including computer program instructions that cause a computerto implement the method of identifying se rvic e s o f ne two rke d resources available for invocation in a distributed object computing architecture as in claims 11 to 15.