1. The Business Process Broker GEOSS Future Products Workshop 26-28 March 2013, Silver Spring, MD, USA Mattia Santoro and Stefano Nativi Institute of Atmosphric Pollution - National Research Council of Italy

2. The BP Broker Use Case A A Create or Edit an abstract BP B B Compile the abstract BP C C Execute the BP D D Return the output to the User USER BP Broker WF engine BP editor Data Publisher

3. Business Process Editor Client Applications which allow users to design Abstract Business Processes (ABP, i.e. technology neutral representation of model interactions) Yaoqiang BPMN 2.0 Editor (Desktop – Web start) Several Eclipse jBOSS plug-in (Desktop) Signavio/Oryx Process Editor (Web-based) Intalio editor BizAgi MS Visio...

4. Workflow Engine Technological frameworks for BP execution. These use a lower level representation of BPs: Executable Business Process (EBP) –jBPM –Express BPEL –Open ESB –OW2 Orchestra –...

5. The Gap Business Processes Design Workflow Frameworks Abstract Business Process These are high-level descriptions, produced by non-IT experts: i.e. modellers, scientist, decision makers, etc. Executable Business Process These are low-level realizations, produced by IT experts ? ?

6. Filling the Gap In an open and heterogeneous context such as the GEO Model Web, an automatic framework for BP transformation needs: –BP Representation; a (set of) well-defined, open specification for BP representation is needed, both for ABPs and for EBPs; –Accessibility: the elements which compose a BP must be accessible on the Web; –Intermediation Services: a set of middle-ware components which provide the necessary intermediation services to harmonize the accessible resources.

7. Intermediation Services: Brokers Web 2.0 “Classic” Geospatial Resources Semantic engines Clients Ontologies/ Gazetteers Resources Image from Stefano Nativi, CNR-IIA (Modified)

8. Intermediation Services: Brokers CSW/ OpenSearch/ OAI-PMH WPS OpenSearch SPARQL Web 2.0 “Classic” Geospatial Resources Semantic engines Clients Ontologies/ Gazetteers Resources Image from Stefano Nativi, CNR-IIA Models Knowledge Bases WF Platforms

9. Business Process Broker A Business Process Broker must provide the needed interoperability arrangements to transform an Abstract BP into an Executable BP. To achieve this, the high-level functionalities provided by a BP Broker are the following: –Discover the needed model implementations in an open, distributed and heterogeneous environment; –Check I/O consistency of BPs and provide suggestions for mismatches resolving: –Publish the EBP as a standard model resource for re-use. –Submit the compiled BP (EBP) to a WF-engine for execution.

10. BP Broker Execution Phases Annotated

11. Pre-processing Steps 1.Components Validation 2.Components compatibility check 1.Mismatches identification 2.Mismatches resolving (through mediation/adaptation approach) 3.Identification of actual I/O parameters of the Executable BP ✗ ✓✓ Annotated

12. Architecture BP Editor Controller BP library Component library Pre processor GEO DAB WF platform BP Broker Knowledge Base Registry

13. Experiments A set of “Use Scenarios” were defined in the FP7 UncertWEB Project –eHabitat, scientific patronage: EC Joint Research Centre (JRC) eHabitat cross-compare –Land-use and Agricultural Policy, scientific patronage: Food and Environment Research Agency (FERA), UK –Activity Modelling, scientific patronage: Technical University of Eindhoven (TUE), NL Deterministic Uncertainty-enabled (2) –Air Quality Forecasting, scientific patronage: Norwegian Institute for Air Research (NILU), NO –Activity Modelling/Air Quality Forecasting Overlay, scientific patronage: University Of Munster (UOM), NILU, TUE

14. Experiments: eHabitat Use Scenario Given a region of interest, the eHabitat model calculates the likelihood to find similar habitats in different regions. The model calculates the Mahalanobis distance using the bio- climatic layers which are used as inputs. The model is developed and published by the EC Joint Research Centre. This scenario implements a Monte Carlo simulation performed on the deterministic eHabitat ecological model. Starting from input datasets with related uncertainty description, the user composes a set of available services to generate a Monte Carlo simulation on the eHabitat model in order to derive a statistics of the output datasets. Output of a run of the scenario using as Protected Area the “Parco Nazionale dello Stelvio” in Italy

15. Experiments: Land-use and Agricultural Policy Use Scenario This scenario is composed of a set of models for predicting land-use and crop yield response to climatic and economic change. There are a total of three models in this scenario, two of which are currently included in the business process. The first model, Land Capability Classification, generates a set of transition matrices describing the likelihood of a field transitioning from one crop to another. The second model, Landsfacts, uses this transition data to simulate field crop allocations over a five-year period. As the Land Capability Classification model generates uncertain transition matrices, an intermediate sampling service is required: Uncertain Transition Matrix Sampler. Output of a run of the scenario, the area of interest is in the neighborhood of York, UK.

16. Activity Modelling/Air Quality Forecasting Overlay Albatross is an agent based model that predicts the schedules of people by predicting which places at which time people visit during a day, what they do in those places, transport modes they use to reach the place, etc. All this information is valuable in connection with emission, exposure, and air quality models. In UncertWEB, the Norwegian Institute for Air Research (NILU) has provided a probabilistic air quality forecast (PAQF) model, which is being used with the simulated activity schedules. This combination provides useful information about how much people are exposed to different types of pollution in time and space. Based on this data, policy makers can develop different policies to decrease the negative effects of low air quality on people. NO2 Exposure Mean NO2 Exposure Standard Deviation NO2 Exposure Mean for Individuals

17. Conclusions: Benefits of the Proposed Approach No need of a composition infrastructure Alleviation from the technicalities of workflow definitions –type matching, identification of external services endpoints, binding issues, etc. User can focus on his/her intended target User may submit incomplete BP definition, and leverage BP recommendations May derive from an aggregated knowledge base of user feedback to execute it (e.g. URR) Of particular interest for multidisciplinary scientific contexts, where different communities may benefit of each other knowledge through model chaining

18. Conclusions: Future Work Future developments will mainly focus on: –Integration of the GEOSS User Requirement Registry –Support of more WF engines –Support of more BP languages –Experiment with new scenarios

19. Thank you for your attention