15/09/2015Semantic Week Page 1 Enhancing Service Selection by Semantic QoS Henar Muñoz Frutos Telefónica I+D
15/09/2015Semantic Week Page 2 Introduction Conclusions and Future Directions Proposed Approach Implementation
15/09/2015Semantic Week Page 3 Context Marketplace model: customers and providers interact for trading services –Introduction of Grid and Cloud technologies –New services included: infrastructure services Services: Storage, computing, infrastructure, software as a Service Provided by Amazon, Google, HP… –Services encapsulate resources Resources: machines, hard disk, applications, licenses Key element: Service Selection based on QoS metrics –Due to the increase of functionally similar services –Due to the importance of non-functional properties: QoS metrics availability, performance, price, reliability, infrastructure properties –To find the best service to meet user requirements and assure QoS
15/09/2015Semantic Week Page 4 Context Important is to have a contract between the involved entities (customer-providers) –Need Service Level Agreement terms for the provision of the service (service description), Quality of service (QoS) level to be maintain, Information about resources to be provided, the liability to compensation if SLAs are not met. SLA templates used for achieving an agreement between customers and provider specifying the terms under the provision of the service SLA templates can be used as input for the service selection based on QoS metrics
15/09/2015Semantic Week Page 5 Problem statement Main SLA specifications –WS-Agreement (OGF-GRAAP)-WSLA (IBM) –They covers the syntactic aspects in the description of SLA: a schema for specifying an agreement and an agreement template, no focus on vocabulary domain –WS-Agreement non-goals: Defining Specification of metrics associated with agreement parameters i.e., how and where these are measured Specification assumes that other complimentary specifications will address these requirements Current SLA specifications lack in interoperability: –Open marketplace: Providers and consumers are collaborating on a global scale competing for offering and consuming services –Different terminologies to describe the vocabulary domain (performance vs. response time) –Different language (performance vs. rendimiento)
15/09/2015Semantic Week Page 6 Example ANSYS request CPUName: IntelCore Duo CPU Speed: 2 GHz Capacity: 400 MB Price: 27 euros per day DiskSpace: 250GB -RAMMemory: 7.5 GB -ComputeUnit: 4 ECU -Storage: 850GB -Platform: 64 bit -Price: $0.4 per instance hour CUSTOMER AMAZON Software engineering company -MemoryPerTask: 7.5 GB -ClockCPUSpeed: 100 MHz / process -StorageCapability: 850GB -Cost: 5 euros/task/hour BSC PROVIDERS Cloud and Grid providers
15/09/2015Semantic Week Page 7 Objectives Improve interoperability between service providers and clients –Providing semantic interoperability by the introduction of semantic annotations –Beyond syntax to semantics, mapping of data exchanged between the involved parties Be backwards compatible –Compatible with current SLA specification e.g. WS-Agreement, WSLA –Compatible with current components (which do not understand the specification…) If specification is used, then extra reasoning is feasible If NOT then the process continues normal operation A lightweight approach, so that, a easy way for providers to introduce these annotations. Increase automation in the selection process to reduce human intervention
15/09/2015Semantic Week Page 8 Conclusions and Future Directions Related Work Implementation Introduction Proposed Approach
15/09/2015Semantic Week Page 9 Proposed Approach Specification for semantic Annotations (SA-SLA) - Backwards compatible - Lightweight approach Conceptual model - QoS ontology Implementation - Semantic Enhanced service Selection (SESS) Selection algorithm - ranking services - taken from [Wang2006]
15/09/2015Semantic Week Page 10 SA-SLA Introduction of semantic annotations in current SLA template files –Based on SA-WSDL –WS-Agreement + WSLA compatible Allow for annotating SLA description with pointers to semantic concepts Annotation of: SLA Parameters. Metrics MetricParameterSLO modelReferencesSchemaMappings Non-semantic descriptions Semantic descriptions RDF, OWL, WSML SA-SLA SLA template QoS metricsSLA terms
15/09/2015Semantic Week Page 11 SA-SLA
15/09/2015Semantic Week Page 12 Conceptual Model Framework capturing the QoS provided by the suppliers and required by the customer to match between the two Conceptual model in QoS which is shared by customers and providers and formalized as OWL –It can be extended by domain requirements (domain specific) QoS ontology: specify SLA parameters and metrics for infrastructure services and their relationships –Based on [Wang2006], [Lee2003], [Tsesmetzis2006], [Ren2007] QoS categorization: infrastructure, system, network, and security and cost
15/09/2015Semantic Week Page 13 QoS ontology
15/09/2015Semantic Week Page 14 Conceptual Model To solve the terminology problem (different terms for the same concept) –Simple Knowledge Organization System (SKOS) –Standard way to represent knowledge organization using RDF SKOS is used for solving terminology problem Response Time Tiempo de Respuesta
15/09/2015Semantic Week Page 15 Conclusions and Future Directions Proposed Approach Introduction Implementation
15/09/2015Semantic Week Page 16 Architecture Comon conceptual model SESS SLA framework SA-SLA (WS-Agreement WSLA) Local domain knowledge Local domain knowledge
15/09/2015Semantic Week Page 17 Ontology Manager Local domain ontology SKOS representation SKOS Manager SA-SLA TSLA Parser Customer Request TSLA Selector SLA template scores SA-SLA Provides Offers Architecture Semantic Enhanced Service Selector
15/09/2015Semantic Week Page 18 Implementation Ontologies –Common conceptual model: OWL –Local knowledge: OWL + RDF (SKOS) Local knowledge repository –Query engine: Jena –Query language: SPARQL –Reasoner: Pellet Semantic Enhanced Service Selector –Parse the SLA templates files (JAXB and XPATH) and –match providers offers with customers’ requirements (selection algorithm taken from [Wang2006]) –Interact with local domain repository for semantic support (create SLA instances, obtain local metrics)
15/09/2015Semantic Week Page 19 Proposed Approach Introduction Implementation Conclusions
15/09/2015Semantic Week Page 20 Conclusions The increasing number of functionally similar services requires non- functional selecting process Non-functional properties as QoS which are described as part of the SLA templates files The introduction of semantic annotations in current SLA specifications allows solving the QoS metric definition problem Using semantic technologies in a non-intrusive way addresses the issue of backwards compatibility and eases development of tools Our proposal is based on existing specifications (WS-Agreement, SAWSDL)
15/09/2015Semantic Week Page 21 References [Wang2006] Wang, X, Vitvar, T., Kerrigan, M., and Toma, I.; A QoS- aware selection model for semantic Web Services. In 4th International Conference on Service Oriented Computing (ICSOC 2006), Chicago, USA, Dec [Lee2003] Lee, K, Jeon, J, Lee, W, Jeong, S., and Park, S., QoS for Web Services: Requirements and Possible Approaches, W3C Working Group Note 25, [Tsesmetzis2006] Tsesmetzis, D. T., Roussaki, I. G., Papaioannou, I. V., and Anagnostou, M. E QoS awareness support in Web-Service semantics. AICT-ICIW. IEEE Computer Society, Washington, 128. [Ren2007] Ren, K.; Jinjun Chen; Tao Chen; Junqiang Song; Nong Xiao; Grid-Based Semantic Web Service Discovery Model with QoS Constraints, Semantics, Knowledge and Grid, Third International Conference on Oct Page(s):479 – 482
15/09/2015Semantic Week Page 22 Thank you for listening Any questions?