Applying 4D ontologies to Enterprise Architecture Matthew West Copyright: SIPC Reference Data Architecture and Standards Manager – Shell Abstract Enterprise Architecture is the term used for a number of related models that together describe an enterprise and its information systems. Many of these models are ontological in nature or have ontological content. A particular problem companies face is that historically many of their models have been "snapshot" or "current state" models and have not been able to manage the history of the enterprise, and of the models used to manage its information. With increasing regulation, such as Sarbanes-Oxley legislation,this is becoming an urgent problem. 4 Dimensionalism is a paradigm that sees objects as extended in time as well as space. As such it has history built in, using a "God's eye view" (outside time). Space-time maps are a technique for showing the patterns different sorts of individuals make in space-time and are a valuable aid to analysis. We will explore the use of these before looking briefly at one or two examples of how they have been used to develop and support Shell's Downstream (oil tanker to petrol pump) data model. 2 What is (an) Ontology? My Answer … Ontology is the study of what exists. An ontology is a theory of what exists. Doesn’t have to be: Formal (computer interpretable) Use a particular form of logic Complete (fully axiomatised) Can be: Philosophical (sorts, identity criteria, lots of discussion) Artificial Intelligence, e.g. First Order or Description Logic based formal ontology. Database structure and data in a database 3 Ontologies 4 How are these ontologies held? • Data Model • Process Model • Locations/Networks • Organization/Workflow • Events/Business cycles • Business Rules • Data Model • Process Model • Data • Documents 5 Ontological Rigour The wider use of Ontologies beyond reasoning Huge potential to apply ontologies to traditional systems Small number of sophisticated applications Vast bulk of information systems Ontology limited or implicit SQL Upper Ontologies Small but increasing number of lower and mid level ontologies Entity OWL DL Relationship OWL Full First Order Logic Language expressiveness Higher Order Logic 6 So what are the practical problems Shell has been grappling with? 7 Differing data models for the same thing in different parts of Shell c1990 Comp. Canada C.I.A. SUKO MF Aus. Model BOSS CMF Arch. Eng. Arch. MF EP SNR SNC Matrix HydroC. DataMdl Ref.Mfg. Data Str. MCSM Area Org'n Loc'n Product Equip. Facility Contract Plan Purch. Prod'n Stock Sale Price Account 8 Data Model Notation: EXPRESS-G entity_y Supertype relationship (subtype at circle end) entity relationship attribute entity_x STRING simple data type 9 2. Cardinalities that lose history • Sometimes cardinalities are set one-to-many (meaning one at a time), when they are really many-to-many because the relationship is transferable. • Imposing restrictions through data structure means: • – Arbitrary or inappropriate restrictions are placed on the data that can be held. – History data about a relationship cannot be held. – Data may be replicated to overcome the restrictions in the data structure. The different versions must be reconciled. – The entity type will only work within the context defined. A change in business rules may require a change in the database structure. – The resultant system is harder to share. This has only become more important with regulations like Sarbanes-Oxley (US accounting rules following the ENRON scandal) 10 Example: Ship registered_at Port Ship registered_under Name Transferable relationships • What happens if you re-register a ship? 11 How can ontology help? 12 3D and 4D approaches to ontology • Data model consistency is dependent on taking a common view of how to represent things across the business. • Unfortunately there are many ways in which we can model the world. • However, there are two main approaches, with on the whole minor variations, that dominate the philosophical literature. • I will call these the 3D paradigm and the 4D paradigm. 13 3D Paradigm • A 3D ontology treats physical objects (roughly things you can kick) as 3D objects (sometimes called continuants) that pass through time. The principles of the 3D paradigm are: 1. Physical objects are 3-dimensional objects that pass through time and are wholly present at each point in time. 2. Physical objects are viewed from the present. The default is that statements are true now. 3. Physical objects do not have temporal parts. 4. Different physical objects may coincide. 14 3D Individual The present (all that exists) space Object passes through time. time 15 4D Paradigm • A 4D ontology treats all individuals – things that exist in space-time - as spatio-temporal extents, i.e. as 4D objects. The principles of the 4D paradigm are: 1. Individuals exist in a manifold of 4 dimensions, three space and one time. So things in the past and future exist as well as things in the present. 2. The four dimensional extent is viewed from outside time rather than from the present. 3. Individuals (including physical objects) extend in time as well as space and have both temporal parts and spatial parts. 4. When two individuals have the same spatio-temporal extent they are the same thing (extensionalism). 16 Possible Individual The past and the future exist as well as the present space Object extended in time time 17 Which paradigm? • The 3D approach corresponds well with the way that language works. Language has a focus around here, now, you and me as a context, and on the current state of affairs. This leads to efficient communication under the most common circumstances. On the other hand dealing with change is relatively problematic. • What is clear is that the 3D and 4D paradigms cannot be merged into a single canonical approach, since they are contradictory, with one requiring physical objects to have temporal parts, and the other forbidding them. • On the other hand, it appears that what can usefully be said using one paradigm can generally be said using the other. • For ISO 15926 (and Shell’s DDM) we chose the 4D paradigm because we found it to be rigorous, and gave a good account of some difficult cases. 18 Space-Time diagrams – an aid to analysis 19 Possible Individual/State – Temporal whole-part space Time period State Individual Events time 20 Materialised Physical Object space event 1 B A event 2 D C Head Handle time The Broom 21 space A game of football – the ball 2nd Half 1st Half Football Match time 22 A game of football – some players Object continues Replaceable Parts Scattered parts space Owen Rooney Player 1 Lampard Player 2 Gerard Player 3 2nd Half 1st Half Football Match time Note: Some replaceable parts are roles 23 A game of football – Roles Replaceable Part/Role Captain space Owen Gerard 2nd Half 1st Half Football Match time Note: Some replaceable parts are roles 24 Your turn to do some work Draw one or more space time diagrams for the participants in the following end-to-end pump impeller replacement activity: • A maintenance engineer requests purchasing to buy a new impeller for a pump. • Purchasing order an impeller from the pump manufacturer sales dept. • The pump manufacturer deliveries department delivers the impeller. • The maintenance engineer replaces the pump impeller. 25 One I prepared earlier – Part 1 End to end impeller replacement process Customer Organization Delivery Impeller Replacement Request Engineering space Purchasing Delivery Sales Order Supplier Organization time 26 One I prepared earlier – Part 2 Impeller Replacement Pump No 3 Impeller space Pump No 3 Engineering time 27 So what does the data model look like? 28 ISO 15926 – Thing 29 Possible Individual whole possible_ individual 9,1 event part composition_of_ individual 1 whole_life_ individual 9,3 participation period_in_time temporal_ whole_part actual_individual 9,2 activity 9,4 temporal_bounding physical_object functional_ physical_object arranged_ individual arrangement_ of_individual (RT) whole 1 assembly_of_ individual materialized_ physical_object stream spatial_location feature_whole_ part 30 Activity (RT) part event (RT) whole (ABS) temporal_bounding caused possible_individual 1 beginning point_in_time ending cause_of_event causer involver activity involvement_ by_reference (RT) whole involved 1,1 thing (RT) part possible_individual participation recognizing recognized recognition 1,1 thing 31 Who is involved in Buying and Selling? Shell’s Downstream Data Model (DDM) has extended ISO 15926-2 from 201 entity types to more than 1700 Many different parties can be involved in buying and selling. Some of these parties are defined in the Organization schema – this is shown on the next slide 32 Buy and Sell parties participation_of_ responsible_ individual agent buyer class_of_participation_ of_responsible_individual (RT) classified_by S[1:?] participating_part_of 2,1 responsible_individual business_role_in_ business_transaction invoicing_party 19,1 class_of_participation_ of_Shell_organization 19,2 participation_of_Shell_ organization 20,1 class_of_supplier stakeholder 17,1 Shell_business_partner 20,2 class_of_customer 21,1 relationship_ administrator shareholder contractor These subtypes give us some of the parties in buying & selling 20,3 trading_party originator inspector 23,1 employee manufacturer_ or_supplier dunning_clerk order_recipient authoriser goods_receipt_clerk 23,5 employer dispatch_clerk purchase_requisition_ authority purchasing_authority 33 Summary • Much business modelling has (or should have) ontology at its core • Ontology has very wide applications – Ontology has been practised by many disciplines – The largest area for the application of ontology is in Business Information Systems • The application of ontology can add considerable value to businesses 34