Presentation is loading. Please wait.

Presentation is loading. Please wait.

A Semantic Peer-to-Peer Overlay for Web Services Discovery

Similar presentations


Presentation on theme: "A Semantic Peer-to-Peer Overlay for Web Services Discovery"— Presentation transcript:

1 A Semantic Peer-to-Peer Overlay for Web Services Discovery
Good, morning, everyone! The title of my presentation is A Semantic Peer-to-Peer Overlay for Web Services Discovery. My name is shuangkai. Shuang Kai

2 Agenda Introduction Thoughtway of our method
Semantic overlay node architecture Routing Table & Neighborhood Table Message Semantic Matching & Routing Example Experimental Evaluation Conclusion This is the outline of my report. First I will give some introduction. And then I will talk about the “Thoughtway of our method”, “Semantic overlay node architecture”, “Routing Table & Neighborhood Table”, “Message Semantic Matching & Routing” in sequence. Finally I will present an example and the Experimental Evaluation of our system . Then conclude my presentation with our future work.

3 Introduction-1 service discovery mechanisms centralized registry
UDDI or DAML-S matchmaker decentralized approach single point failure performance bottleneck In recent years, more and more Web Services are becoming available on the Internet. The growing number of Web Services demands for a scalable, flexible and reliable solution to discovery the most appropriate services for the requesters. The mechanisms of service discovery include centralized registry and decentralized approach. Much of the work on Web Services discovery is based on the centralized registries, like UDDI or DAML-S matchmaker. However as the number of Web Services grows and become more dynamic, the centralized registries quickly become impractical because of the single point failure and performance bottleneck.

4 Introduction-2 based on P2P technology unstructured P2P network
limitation on the scalability structured P2P networks based on DHT logn routing hops – network of size n nodes exact match In order to avoid the disadvantages of the centralized systems, a number of decentralized solutions based on P2P technologies have been proposed. There are two main directions, one direction is based on unstructured P2P network, and the other is based on structured P2P network. But the unstructured P2P network based method has much limitation on the scalability. The structured P2P networks based on Distributed Hash Table (DHT) have very good properties for scalability. And the lookups can be resolved in logn routing hops for a network of size n nodes. But current methods only support exact match lookups and encounter difficulties in complex queries.

5 Agenda Introduction Thoughtway of our method
Semantic overlay node architecture Routing Table & Neighborhood Table Message Semantic Matching & Routing Example Experimental Evaluation Conclusion

6 Thoughtway of our method
structured P2P semantic method different matching degree extends the Plaxton mesh dynamic semantic overlay network keyword prefix routing to semantic prefix routing So we present a structured peer-to-peer semantic routing architecture, which could efficiently route and locate the semantic service request to service registration node. And our method support different matching degrees. We extended the Plaxton mesh to dynamic semantic overlay network to manage service advertisements and lookups. We extended prefix routing to semantic prefix routing. The approach makes it possible to quickly identify the peers containing most likely matched services.

7 Characteristic Vector - 1
semantic service description OWL-S or WSMO characteristic vector a series of numeric string extract service’s information input, output, Pre-conditions and Effects ontological concept A semantic service description, including service advertisements and service requests can be described in service ontology, like OWL-S or WSMO. We extract the information from semantic service description, such as input, output, Pre-conditions and effects as ontological concept. All those ontological concepts together represent the service. We encode these ontological concepts to numeric elements, called the service’s characteristic vector, acronym CV. So CV is a series of numeric strings, one numeric string represent a concept. In this paper, we used binary string.

8 Characteristic Vector - 2
First we model the ontology graph as a multibit-trie where each ontological concept is a node. For each node of the trie, we label each branch in a binary string. Then we define the key of the ontological concept as a string composed by the branch labels from root to the node representing the concept. This encoding scheme make the code of concepts have the prefix property, which represent the hierarchical relationship among ontological concepts. Suppose there is a service description, Advertisement1 = {C7, C6, C5, C4}, the CV of this description is CV={10, 11, 0000, 0100}. In each layer of the trie, the elements of CV sort in ascending order or descending order . service description: Advertisement1 = {C7, C6, C5, C4} CV={10, 11, 0000, 0100}.

9 Agenda Introduction Thoughtway of our method
Semantic overlay node architecture Routing Table & Neighborhood Table Message Semantic Matching & Routing Example Experimental Evaluation Conclusion

10 Semantic overlay node architecture -1
This slide shows the structure of a single overlay node. The execution and communication module provides an information bus to connect the other internal components. It interacts with external parties, i.e., users, service providers, and other peers to get service message and provide these information to the internal components. The query processor transforms service semantic description to characteristic vector which could be handled by our system. The characteristic vector of service is then sent to the matchmaker, and the matchmaker with knowledge base will complete ontology-based semantic matching between service query request and service advertisements. With the match result returned by matchmaker, decision was made which peer the message will be forwarded to, or return the service information stored in this peer.

11 Semantic overlay node architecture -2
Each peer also maintains a routing table and a neighborhood table, these tables contains entries of the next hop. These tables are extended from those of the Plaxton mesh.

12 Agenda Introduction Thoughtway of our method
Semantic overlay node architecture Routing Table & Neighborhood Table Message Semantic Matching & Routing Example Experimental Evaluation Conclusion

13 Routing Table multiple rows each rows holds a number of entries
semantic matching of the prefix up to an element in the CV A node routing table has multiple rows, where each rows holds a number of entries. Each entry in the routing table contains the IP addresses of that peer. Each row represents a semantic matching the prefix up to an element in the characteristic vector. These tables are extended from those of the Plaxton mesh and has similar meaning.

14 Neighborhood Table Contains the nodeID and IP address of the peers
Set of peers that are closest to the local peer Based on semantic similarity The neighborhood table, contains the nodeID and IP address of the peers, is a set of peers that are closest in proximity to the local peer. The proximity that the presented system uses is based on semantic similarity between the peers and the local peer.

15 Agenda Introduction Thoughtway of our method
Semantic overlay node architecture Routing Table & Neighborhood Table Message Semantic Matching & Routing Example Experimental Evaluation Conclusion

16 Semantic Matching Degree
Exact If advertisement S and request Q are equivalent S ≡ Q PlugIn If S could always be used for Q S is a plug-in match for Q, Q⊆S. Subsume If S is more general than Q S is a Subsume match for Q, S⊆Q In our method, there are three kinds of results from matchmaking between service request Q and service advertisement S: Exact, PlugIN, Subsume.

17 Message Routing and Locating
When a peer receives a message, it first converts the service description to its characteristic vector (CV), and calculates the semantic similarity between CV and peers in the neighborhood table to see if the CV falls within the range of semantic distance covered by the neighborhood table. If so, the message is forwarded directly to the destination node. If the neighborhood table does not satisfy the condition of semantic distance with the request, then the routing table will check the entries that share a common prefix with the CV by at most elements. The service message can be forwarded to the node satisfied “Exact” query. Meanwhile, the service message will be forwarded to the nodes are found to satisfy “Subsume” or “PlugIn” query under the users preference.

18 Agenda Introduction Thoughtway of our method
Semantic overlay node architecture Routing Table & Neighborhood Table Message Semantic Matching & Routing Example Experimental Evaluation Conclusion

19 Example Message ID C3C7CAC0 Subsume Exact 2007-1-5
This figure gives an example of semantic overlay routing. Here, we can see that the message with ID C3C7 CAC0, the message path begins from node C6C3C0CC. According to the matching request, node C3C7CBC2 forwards the message to two nodes: C3C7CAC2 and C3C7C5C1.

20 Agenda Introduction Thoughtway of our method
Semantic overlay node architecture Routing Table & Neighborhood Table Message Semantic Matching & Routing Example Experimental Evaluation Conclusion

21 Experimental Evaluation - 1
We evaluate the routing performance of our system by measuring average search hops and the distribution of the number of routing hops. We generate randomly characteristic vectors which are uniform distribution. The first experiment makes a comparison between the presented system and Pastry in the number of routing hops. We measure the average search hops in the network with different numbers of nodes.

22 Experimental Evaluation - 2
The second experiment evaluates the distribution of the number of routing hops. The actual number of nodes in our system can be much more than the number shown in the graph. Thus we conclude that out system is both scalable and efficient in terms of discovery. Compared with a centralized service discovery system, our system is more scalable.

23 Agenda Introduction Thoughtway of our method
Semantic overlay node architecture Routing Table & Neighborhood Table Message Semantic Matching & Routing Example Experimental Evaluation Conclusion

24 Conclusion Our System Future work support semantic matching
ontological concept encoding scheme based on the structured overlay network dynamic Plaxton mesh-like network Future work improve the usability of this system From the above, we can see that our system can support semantic matching based on ontological concept encoding scheme, and on the structured overlay network, which is extended from Plaxton mesh network. And our future work will focus on improving the usability of our system.

25 Thank you!


Download ppt "A Semantic Peer-to-Peer Overlay for Web Services Discovery"

Similar presentations


Ads by Google