Business Processes as Artifacts Jianwen Su University of California, Santa Barbara Nanjing U/2012 Summer School 2012/07/24
The “Big Data” Report n Mckinsey Global Institute, June 2011: Big data: The next frontier for innovation, competition, and productivity n MGI: established in 1990 to develop deeper understanding of the evolving global economy Mission: To provide leaders in the commercial, public, and social sectors with the facts and insights on which to base management and policy decisions Nanjing U/2012 Summer School 2012/07/24 2
From EXECUTIVE SUMMARY: “The United States alone faces a shortage of 140, 000 to 190, 000 people with deep analytical skills as well as 1. 5 million managers and analysts to analyze big data and make decisions based on their findings. ” Nanjing U/2012 Summer School 2012/07/24 3
What big data can generate: Nanjing U/2012 Summer School 2012/07/24 4
Business (Biz) Processes n. A biz process is a set of one or more linked activities (automated or manual) that collectively realize a business objective or policy goal, normally within the context of an organizational structure defining functional roles and relationships Obtaining a Permit Nanjing U/2012 Summer School 2012/07/24 5
BP Management Systems (BPMSs) Software systems to manage and support (and control) n biz models n data (documents, files, …) BPM system n enactments n resources (including human) n others (e. g. auditing) BP “=” workflow in the wider sense Traditional meaning of workflow in 80’s to early 90’s means task sequencing Nanjing U/2012 Summer School 2012/07/24 6
Outline n Challenges in Business Process Management n Artifact-centric n EZ-Flow Modeling Approach and Selected Technical Issues n Conclusions Nanjing U/2012 Summer School 2012/07/24 7
Vanda Group AVT n Developing workflow systems for regional banks, credit unions, provident funds, … n Est. 60% of the market excluding national banks Key obstacles: n Training (engineer liquidity) n Repetition of work, labor intensive (could make more $$ or ¥¥ and be more competitive) n High maintenance cost Nanjing U/2012 Summer School 2012/07/24 Vanda (1982 -) developed workflow application domains 8
Hangzhou Housing Management Bureau n Population: 8. 7 millions One division (~400 SMEs) deals with all real estate licenses, permits, titles, etc. n 300, 000 cases each year, ~500 workflow (types), 35% 1 day, 30% 7 -9 days developing workflow application domains Nanjing U/2012 Summer School 2012/07/24 9
Hangzhou Housing Management Bureau n Population: 8. 7 m One division (~400 SMEs) deals with all real estate licenses, permits, titles, etc. n 300, 000 cases each year, ~500 workflow (types), 35% 1 day, 30% 7 -9 days [Jin-Wen-Wang Coop. IS 2011] reference models: 600+ 3, 000+ 6, 000+ 200, 000+ Nanjing U/2012 Summer School 2012/07/24 10
Hangzhou Housing Management Bureau n Population: 8. 7 m One division (~400 SMEs) deals with all real estate licenses, permits, titles, etc. n 300, 000 cases each year, ~500 workflow (types), 35% 1 day, 30% 7 -9 days n Contractor/in-house development of workflow system(s) (¥¥ millions for in-house only) Challenges: n Manage changes (policy, environment, …) n Serious lack of automation for design-development-maintenance developing workflow application domains Nanjing U/2012 Summer School 2012/07/24 11
Hospitals: Rui. Jin & Cottage n Health 上海瑞金医院 care delivery: much of the $300 billion could be gained n Treatment workflows can fundamentally improve health care quality Falling far behind: n No workflows, conflicting “workflows” n “Shaky” IT infrastructures new IT divide? n Rui. Jin has the largest IT team (40+FTEs) among all hospitals in Shanghai wishful workflow application domains Nanjing U/2012 Summer School 2012/07/24 12
Application and Research Challenges n Lack of clear ways to combine various factors of workflows n Lack of workflow technology to support a variety of essential functions n Long tail phenomenon is a “holy grail” n Application domains work in isolation n Unifying holistic conceptual models n Design and runtime support n Reasoning, business “informatics”, process mining n Interoperation Nanjing U/2012 Summer School 2012/07/24 13
Outline n Challenges in Business Process Management n Artifact-centric n EZ-Flow Modeling Approach and Selected Technical Issues n Conclusions Nanjing U/2012 Summer School 2012/07/24 14
The Challenge of BPM Business Strategy • “Be more green” • “Use our differentiators” High Manager Business Architect Solution Designer Nanjing U/2012 Summer School High Executive Business Goals Business Architecture Business Optimization 2012/07/24 15
A Representative “Model” at Biz Manager Level A Business Component Map is a tabular view of the business components in the scope of interest “Accountability Level”: scope and intent of activity and decision-making t ”: n ne rise l po rp tia om nte ten C ss of e s po ineart ha ate ntly s Bu p hat per nde “ t o pe to de in “Business Competencies”: large biz area with characteristic skills and capabilities Business Administrati on directing Business Planning controllin g Business Unit Tracking Staff Appraisals executin g Nanjing U/2012 Summer School Staff Administrati on Production Administrati on New Business Developme nt Sector Planning Sector Manageme nt Product Manageme nt Relationshi p Manageme nt Account Planning Relationship Managemen t Credit Assessment Servicing & Sales Product Fulfillment Financial Control and Accounting Sales Planning Fulfillment Planning Portfolio Planning Marketing Campaigns Credit Administrati on 2012/07/24 Fulfillment Planning Sales Product Directory Sales Manageme nt Product Fulfillment Customer Dialogue Contact Routing Document Manageme nt Compliance Reconciliati on Customer Accounts General Ledger 16
The Challenge of BPM Business Strategy High Executive • “Be more green” • “Use our differentiators” Business Goals Business Architecture Business Optimization High Manager Business Architect Solution Designer Customers Employees Business Operations Partners Resources IT Nanjing U/2012 Summer School 2012/07/24 17
Common Model at IT Level: An Activity Flow is a (typically) graph-based specification of how activities/processes are to be sequenced Process Modeling Business Logic Direct, flow-based implementation Data Modeling System in Operation Nanjing U/2012 Summer School Biz Process Management System (flow mgmt, services, databases, resources, …) 2012/07/24 18
The Challenge of BPM Business Strategy High Executive • “Be more green” • “Use our differentiators” Business Goals Business Architecture Business Optimization High Manager Business Architect Solution Designer Customers Employees Partners Speak in terms of 4 “Functional Decomposition” 4 “Business Components” Hard to Communicate !! Business Operations Resources IT Nanjing U/2012 Summer School Operations need to be v Faithful v Measurable v Flexible 2012/07/24 Speak in terms of 4 “Workflow” 4 “Process centric” 4 “Activity-flow” 19
Common Model at IT Level: An Activity Flow is a (typically) graph-based specification of how activities/processes are to be sequenced Process Modeling Business Logic Direct, flow-based implementation Data Modeling System in Operation Biz Process Management System (flow mgmt, services, databases, resources, …) n Data and business objects are typically an afterthought n Hard for stake-holders to communicate about the big picture v People “see the trees but not the forest” v Overall process can be chaotic – Cf. “staple yourself to a customer order” n Hard to manage versions v E. g. , evolution, re-use, generic workflow with numerous specializations Nanjing U/2012 Summer School 2012/07/24 20
Typical Biz Process Modeling n. A bookseller example: Traditional control-centric models Fill Shopping Cart ID Customer Nanjing U/2012 Summer School Shipping Preference Payment information 2012/07/24 Confirmation Archive 21
Typical Biz Process Modeling n. A bookseller example: Traditional control-centric models n Multiple steps needed for each activity Fill Shopping Cart ID Customer Shipping Preference Payment information Confirmation Archive Ground Credit Air In-stock Handling Check Inventory Back-order Handling Warehouses/ Size New Existing Customer Registration Login Pay. Pal Check In practice, 100 s to 1000 s of nodes Hard to reason, find useful views: missing data Nanjing U/2012 Summer School 2012/07/24 22
BP Analytics (Biz Intelligence) n Extract-Transform-Load inventory Transactions catalog Transactions activities Data Warehouse Analysis Biz Process is missing! Transactions Nanjing U/2012 Summer School cust_db 2012/07/24 23
Why We Should Look for a Unifying Model Good models go beyond description – they support action n Selecting the right model for the job matters Example: “Game of 15” Winner: First one to reach exactly 15 with any 3 chips 1 2 3 First model – A is 4 5 and B is Second model – 6 7 8 9 – what is B’s move? – B’s move is 6! Example due to a “model”(IBM) Can we find David Cohn of business operations that is • Useful & natural for the business level stake-holders to use • Useful & natural for mapping to 2012/07/24 infrastructure the IT Nanjing U/2012 Summer School 24
Data Management In the Infancy (60’s) n Driving applications: inventory control, financial data management query By hand COBOL program n The logical data model Labor intensive File structures (indexes, …) desirable have to deal with key to the success: automation Nanjing U/2012 Summer School 2012/07/24 25
A Fundamental “Theorem” of Databases n Physical data independence allows us to focus only data management issues SQL logical data model conceptual query plan Nanjing U/2012 Summer School physical organization (files, pages, indexes, …) 2012/07/24 26
Future of BPM? n Automate ’s changes Changes to system process model data model system (model) (databases, services, workflows, resources) business IT n Reuse concepts, tools, techniques developed in CS n First step: a single conceptual model for biz processes v both data and processes are 1 st class citizens Nanjing U/2012 Summer School 2012/07/24 27
Outline n Challenges in Business Process Management n Artifact-centric n EZ-Flow Modeling Approach and Selected Technical Issues n Conclusions Nanjing U/2012 Summer School 2012/07/24 28
BP Modeling: Data Exclusion to Data Centricity n Data exclusive models focus on activity flow and management v Wf. MC, BPMN, … n Incorporating data as views complements well (but separate from) activity views v UML (object modeling and activity diagrams) n Executable models integrate data and activities with low level of abstraction v BPEL n Recent data-centric approaches treat both data and activities “equally” in a more uniformed manner v Biz artifact-centric, form-based, spreadsheet-based Nanjing U/2012 Summer School 2012/07/24 29
Business Artifacts n. A business artifact is a key conceptual business entity that is used in guiding the operation of the business v fedex package delivery, patient visit, application form, insurance claim, order, financial deal, registration, … v both “information carrier” and “road-maps” n Very natural to business managers and BP modelers n Includes two parts: v Information model: data needed to move through workflow v Lifecycle: possible ways to evolve Nanjing U/2012 Summer School 2012/07/24 30
Example: Restaurant Artifacts Activity Create Guest Check Kitchen Order repository Add Item Open GCs Prepare Receipt Pending KOs Cash Balance Pending Receipts Payment Prepare & Test Quality Closed GCs Paid Receipts Ready KOs Update Cash Balance Deliver Disagreed Receipts Recalculate Receipt Nanjing U/2012 Summer School Archived Receipts Archived GCs 2012/07/24 Cash Balance Archived KOs 31
Example: Restaurant Artifacts Create GC Guest Check Kitchen Order KO Add Item Open GCs Prepare Receipt RC Receipt Pending KOs Cash Balance Pending Receipts Payment Prepare & Test Quality Closed GCs Paid Receipts Ready KOs Update Cash Balance Deliver Disagreed Receipts Recalculate Receipt Nanjing U/2012 Summer School Archived Receipts Archived GCs 2012/07/24 Cash CB Balance Archived KOs 32
![Case Study : IBM Global Financing [Chao, Cohn, et al BPM 2009] n Finance](https://present5.com/presentation/28bdfd733e9940789691a1cd791a3390/image-33.jpg "Case Study : IBM Global Financing [Chao, Cohn, et al BPM 2009] n Finance")
Case Study : IBM Global Financing [Chao, Cohn, et al BPM 2009] n Finance HW, SW & services from IBM & others for clients n IBM internal financing business w/ global reach v World’s largest IT financier w/ $38 B asset base v Financing >$40 B IT assets / year for last 3 years v 125 K clients across >50 countries (9% of IBM profit) n Business challenges v Operations tailored to mega-deals becoming too costly v Efficiency & cost control required global performance metrics v Country “silos” inhibited integration & annoyed clients v Current methods failed to produce end-to-end “tangible model” v Needed globally standard process w/ local variations Nanjing U/2012 Summer School 2012/07/24 33
How the Artifact-Centric Approach Helped n In a 3 -day workshop with 15 business SMEs from IGF, a preliminary artifact design was created v Already useful to stakeholders from different regions as a common vocabulary n 6 weeks of design refinements lead to final design v Enabled visibility into the global process and the regional variations: not possible before v A blueprint for transformation of IGF operations l VP roles assigned to pieces of top-level artifact model n Current plan: automate the global-level artifact model v Anticipate significant improvement in efficiency v Plan to substantially augment the sales staff Nanjing U/2012 Summer School 2012/07/24 34
Emerging Artifact-Centric BPs customer info cart . . . + Specification of artifact lifecycles Artifacts (Info models) n Informal model [Nigam-Caswell IBM Sys J 03] n Systems: BELA (IBM 2005), Siena (IBM 2007), Arti. Flow (Fudan-UCSB 2010), Barcelona (IBM 2010) n Formal models v State machines [Bhattacharya-Gerede-S. SOCA 07][Gerede-S. ICSOC 07] v Rules [Bhattacharya-Gerede-Hull-Liu-S. BPM 07][Hull et al WSFM 2010] Nanjing U/2012 Summer School 2012/07/24 35
Outline n Challenges in Business Process Management n Artifact-centric n EZ-Flow Modeling Approach and Selected Technical Issues n Conclusions Nanjing U/2012 Summer School 2012/07/24 36
Artifact-Centric BPMSs @IBM: n Declarative models n Semantics (U Rome) n Analysis (UCSD) n Workflow views (lenses) Conceptual BP models Optimization execution control @UCSB in collaboration with IBM, U Rome, Fudan, … Nanjing U/2012 Summer School BPMS components 2012/07/24 37
Declarative Biz Processes + Artifacts (info models) n Variation of Nanjing U/2012 Summer School + Semantic services (IOPEs) if C enable … Conditionaction rules [Bhattacharya-Gerede-Hull-Liu-S. BPM 07] 2012/07/24 38
EZ-Flow: Procedural Biz Processes n Each biz process has a core artifact (class) v Business data (object) + enactment v Event driven v Similar notion in recent GSM model from IBM [EZ-Flow or Arti. Flow, 2009, 2010] Nanjing U/2012 Summer School 2012/07/24 40
EZ-Flow Engine … e 3 e 1 event queue EZ-Flow Scheduler e 2 exec(T 2, PAF 01) perform T 2 exec(T 3, PAF 05) perform T 4 perform T 3 task performer: handles data wrapping and service wrapping Nanjing U/2012 Summer School 2012/07/24 41
EZ-Flow and Research Problems Nanjing U/2012 Summer School 2012/07/24 42
EZ-Flow and Research Problems verification dominance automated construction preserve data ICs runtime monitor dynamic modification process ICs exec. res. calculation Nanjing U/2012 Summer School 2012/07/24 43
EZ-Flow and Research Problems verification dominance automated construction preserve data ICs runtime monitor dynamic modification process ICs exec. res. calculation Nanjing U/2012 Summer School 2012/07/24 44
Changes in Biz Processes n Reason for changes: v Policy/regulation change v Technology change v Environment change v User demand change v… n The long tail phenomenon: large number of cases of a small number of patterns a small number of cases are mostly different n BPMSs must handle the latter more efficiently Nanjing U/2012 Summer School 2012/07/24 45
Manage Changes n Modify biz process model: time consuming, big effort n Anticipate change at design time, and build flexibility in schema, e. g. , [Gottschalk-van der Aalst-Jansen-Vullers-La Rosa 2008] [Hallerbach-Bauer-Reichert 2008] v limited options n Declarative models: worklet [Adams-ter Hofstede-Edmond-van der Aalst 2006], LTL-based [van der Aalst-Pesic-Schonenberg 2009] v Data not included n Runtime dynamic execution mechanism based on objects (task wrappers) [Redding-Dumas 2010] v Detached from process model, low abstraction n Our approach: procedural process model with declarative changes, conservative extension [Xu-S. -Yang-Zhang 2011] Nanjing U/2012 Summer School 2012/07/24 46
Technical Approach n Ingredient n Each 1: artifact-centricity biz process has a core artifact (class) Nanjing U/2012 Summer School 2012/07/24 47
Technical Approach n Ingredient 2: formal model (semantics) for execution n Ingredient 3: declarative change specification v Four execution altering operators v Rules for applying the operators based on conditions Nanjing U/2012 Summer School 2012/07/24 48
Natural Disaster Victims on Green Channel Express-SR: MAY skip Secondary. Review ON PAF WHERE project. Type="resettled" Nanjing U/2012 Summer School 2012/07/24 49
New Fee Schedule for Low Incoming Housing Affordable-Fee: MUST REPLACE Payment. Processing BY Affordable. Payment. Processing ON PAF WHERE SELF. project. Type="affordable" Nanjing U/2012 Summer School 2012/07/24 50
New Contractor Needs Prequalification First-Timer: MUST ADD Prequal BEFORE Prelim_Decision ON PAF WHERE project. Type="affordable" AND developer. Name NOT IN SELECT developer. Name FROM PAF P WHERE P. artifact. Id <> SELF. artifact. Id AND P. project. Type="affordable" Nanjing U/2012 Summer School 2012/07/24 51
Insufficient Selling Space Need Re-Check Re-eval: MUST RETRACT FROM Secondary. Review TO Receiving. App-form ON PAF WHERE SELF. cp. plan. Area < ( SELECT sum(P. selling. Area) FROM PAF P WHERE P. cp=SELF. cp GROUP BY p. artifact. Id ) Nanjing U/2012 Summer School 2012/07/24 52
Mixed Procedural and Declarative Pays off n Biz process = state machine lifecycle + change rules n Modification rules conservatively extend workflow v Could be temporary, non-schematic n Allows biz process to respond to situations with many more options: v # of “trace types” grow exponentially in # rules n Performance estimates: v 9% labor savings for Real Estate Administration of Hangzhou (preliminary study) [Xu-S. Yan-Yang-Zhang 2011] Nanjing U/2012 Summer School 2012/07/24 53
EZ-Flow and Research Problems verification dominance automated construction preserve data ICs runtime monitor dynamic modification process ICs exec. res. calculation Nanjing U/2012 Summer School 2012/07/24 54
Why Comparison Many reasons: n Optimization (similar to comparing queries) n Replacing part of workflow (reorganization) n Updating workflow (evolution) n Reusing workflow n. . . Nanjing U/2012 Summer School 2012/07/24 56
Workflow Dominance A S y t porar ut tpu inp ou tem A R if C enable … S ary r t ut utpu empo inp o t R if C enable … W 1 W 2 if every input-output pair that can be produced by W 1 can also be produced by W 2 n Note: v their temporary data can be very different v services are different; rule sets are different v services may be done by human Nanjing U/2012 Summer School 2012/07/24 57
Performance Policies performance policy p is a function that assigns each service s a multi-valued function over U n. A A s B p(s) : x x+1, x+2 n Since the “flow” is fixed, the choice of a performance policy determines how the workflow would perform v E. g. , given an input, a workflow can execute and generate an output P of performance policies p v Absolute (ABS): p(s) = U U v Fixed choice: p(s) is some single-valued function n Classes Nanjing U/2012 Summer School 2012/07/24 58
![Definition of (k-)Dominance n Fix [Calvanese-De Giacomo-Hull-S. ICSOC 09] a class of performance policies](https://present5.com/presentation/28bdfd733e9940789691a1cd791a3390/image-57.jpg "Definition of (k-)Dominance n Fix [Calvanese-De Giacomo-Hull-S. ICSOC 09] a class of performance policies")
Definition of (k-)Dominance n Fix [Calvanese-De Giacomo-Hull-S. ICSOC 09] a class of performance policies P A S ary t tput por u inp ou tem A R y t porar ut tpu inp ou tem if C enable … k W 1 P S R if C enable … W 2 if for each performance policy p 1, there is a performance policy p 2, such that every input-output pair produced by W 1[p 1] in at most k steps can also be produced by W 2[p 2] in at most k steps Nanjing U/2012 Summer School 2012/07/24 59
Results on (k)-Dominance n Absolute k-dominance is decidable but dominance is undecidable: 1. (Z, +, <), integers with additions 2. (Q, +, <), rational numbers with additions 3. (R, +, × , <), real numbers with additions and multiplications (the real closed field) n Absolute dominance is undecidable: 1. (Z, <), integers with discrete order 2. (Q, <), rational numbers with dense order 3. (R, <), real numbers with dense order [Calvanese-De Giacomo-Hull-S. ICSOC 09] Nanjing U/2012 Summer School 2012/07/24 61
EZ-Flow and Research Problems verification dominance automated construction preserve data ICs runtime monitor dynamic modification process ICs exec. res. calculation Nanjing U/2012 Summer School 2012/07/24 62
Synthesis Problem n Given a goal and a set of services, construct a set of rules so that every execution satisfies the goal + Artifact (Info model) Semantic services (IOPEs) + Goal (FO) ? if C enable … Conditionaction [Fritz-Hull-S. ICDT 09] (restricted to single artifact, first-order goals) Nanjing U/2012 Summer School 2012/07/24 63
Artifact Schema n An artifact schema is a finite set A of attributes. An artifact of A is a mapping from A to U n Assume a set of initial attributes Ainit A artifact is B-completed, B A, if it is defined on all attributes in B v “input” artifacts are Ainit-completed n An Nanjing U/2012 Summer School 2012/07/24 64
Semantic Services (Tasks) n. A semantic service over A is a tuple (s, R, W, p, r), where vs : service name v R, W : finite sets of (resp. , read, write) attributes v p, r : quantifier-free formulas (pre- and post-condition, resp. ) over R, R W, resp. n allow DEF(A) for an attribute A is the result of executing s on o, o o , if s (o, o ) = p r, and v n o v frame conditions are satisfied Nanjing U/2012 Summer School 2012/07/24 65
An Example Semantic Service A 0 A 2 Nanjing U/2012 Summer School s B 0 A<1 0 B 1 A 2 1 B 2012/07/24 66
Condition-Action Rules condition-action rule is an expression “if enable s” where v is a (quantifier-free) formula and v s is a semantic service n. A is the result of executing a rule r : if invoke s on o, o r o , if v o = , and v o o s n o Nanjing U/2012 Summer School 2012/07/24 67
Workflow Schema workflow schema is a triple W = (A, S, R) v A : artifact schema v S : a finite set of semantic tasks v R : a finite set of condition-action rules n. A n Denote * the closure of r Nanjing U/2012 Summer School r R 2012/07/24 68
![The Synthesis Problem [Fritz-Hull-S. ICDT 2009] n Given A, Ainit, a finite set S](https://present5.com/presentation/28bdfd733e9940789691a1cd791a3390/image-66.jpg "The Synthesis Problem [Fritz-Hull-S. ICDT 2009] n Given A, Ainit, a finite set S")
The Synthesis Problem [Fritz-Hull-S. ICDT 2009] n Given A, Ainit, a finite set S of semantic tasks, a formula over Afinal A, Find a rule set R such that * if o o , then o = Ainit-completed Afinal-completed S A + Artifact (Info model) Nanjing U/2012 Summer School + Semantic services (IOPEs) 2012/07/24 Goal (FO) ? R = Conditionaction 69
A Trivial Solution S A + Artifact (Info model) n Just + Semantic tasks (IOPEs) Goal (FO) ? R = Conditionaction let R = n Need to revise the problem statement Nanjing U/2012 Summer School 2012/07/24 70
Maximally Safe Ruleset n. A ruleset enables all executions that guarantee to satisfy the goal n Goal: 1 B A 0 A 2 s B 0 A<1 0 B 1 A 2 1 B A 2 : definitely good 0 A < 1 : possibly good but can’t be sure n Best we can do n 1 if 1 A 2 enable s Nanjing U/2012 Summer School 2012/07/24 71
Maximally Safe Ruleset With Exception n. A ruleset eagerly move dead-end executions to EXCEPTION status n Goal: 1 B A 0 A 2 s B 0 A<1 0 B 1 A 2 1 B A 2 : definitely good 0 A < 1 : possibly good but can’t be sure n Be optimistic: n 1 if 0 A 2 enable s if B < 1 goto EXCEPTION Nanjing U/2012 Summer School 2012/07/24 72
Pre-Conditions a semantic task (s, R, W, p(x), r(xy)), and a (subgoal) condition d(xy) n A -precondition of s, d is a formula e(x) such that v e logically implies p and r p v x (e(x) ( y r(xy) d(xy)) holds s WP (s, d) : weakest -precondition n Given -precondition of s, d is a formula e(x) such that v e logically implies p and v x (e(x) ( y r(xy) d(xy)) holds WP (s, d) : weakest -precondition n. A Nanjing U/2012 Summer School 2012/07/24 73
Weakest Pre-Conditions a semantic task (s, R, W, p(x), r(xy)), and a (subgoal) condition d(xy) n The weakest -precondition WP (s, d) p(x) ( y r(xy) d(xy)) n Given useful for maximally safe ruleset n The weakest -precondition WP (s, d) p(x) ( y r(xy) d(xy)) useful for maximally safe ruleset with exception Nanjing U/2012 Summer School 2012/07/24 74
Necessary Condition Theorem: If there exists an algorithm to find maximally safe rule sets, the FOL theory is decidable (for the context structure) Nanjing U/2012 Summer School 2012/07/24 75
The Other Direction n Invoke-once constraint: each semantic task is allowed to run once Theorem: Under the invoke-once constraint, if the FOL theory (of the structure) is decidable and admits quantifier elimination, then the maximally safe rule sets can be computed Nanjing U/2012 Summer School 2012/07/24 76
A Special Case: Dense Order (Q, <) n Goal and task conditions are quantifier free formulas n Acyclic task invocation dependencies n Each task writes one attribute Theorem: Computing Maximal Safe Ruleset is PSPACE-complete n Key ideas: cell decomposition; reduction from QBF n Acyclicity condition can be dropped [Hull-S. 2009] (in preparation) Nanjing U/2012 Summer School 2012/07/24 77
Further Restrictions n. A constructive EXPTIME algorithm n PTIME if #needed attributes is bounded Nanjing U/2012 Summer School 2012/07/24 78
![Summary of Results [Fritz-Hull-S. ICDT 2009] n Synthesis problem is harder than FO logic](https://present5.com/presentation/28bdfd733e9940789691a1cd791a3390/image-76.jpg "Summary of Results [Fritz-Hull-S. ICDT 2009] n Synthesis problem is harder than FO logic")
Summary of Results [Fritz-Hull-S. ICDT 2009] n Synthesis problem is harder than FO logic theory of the underlying structure n Positive answer for special cases v Invoke once v Concrete algorithm for dense order domain: PSPACE-complete Nanjing U/2012 Summer School 2012/07/24 79
EZ-Flow and Research Problems verification dominance automated construction preserve data ICs runtime monitor dynamic modification workflow ICs exec. res. calculation Nanjing U/2012 Summer School 2012/07/24 80
An Example Workflow – Ez. Mart Environment (customer, manager, …) Register request Checkout Customer register Pay by bank Bank reply Order pay Order create Order paid Contact customer support Customer support reply Order further action Ship prepare Order Inventory action taken sell … … … n Traditional workflow specifications v Centered on control flow v Data flow is embedded in workflow executions [X. Liu-S. -Yang, 2011] Nanjing U/2012 Summer School 2012/07/24 81
Data and constraints Customer Order custid email UNIQUE addr name ordid Inventory invid prod qtyav loc custid NOT NULL invid NOT NULL shipid qty ostat Ship shipid ordid NOT NULL addr NOT NULL name NOT NULL from NOT NULL status Data integrity constraints In data schema v key, foreign key, candidate key UNIQUE, not-null n On attribute content v Order: qty>0; Ship: from addr n Business specific constraints v Status: order cannot be canceled or returned when there is an associated shipment not finished v… n n Nanjing U/2012 Summer School 2012/07/24 82
GSM: A Declarative Workflow Language Customer + registered checkout Order payorder ordid=payorder. ordid … + created pay paid ostat : =CREAT; qty : = checkout. qty; custid : = checkout. custid; . . . Ship + prepare ready paid+ … Inventory + Nanjing U/2012 Summer School inv initiated ship 2012/07/24 custsuppreply. ostat further actiontaken action invokes custsupp sent paid+ … sell ostat : = sold deliver report result qtyav<10 update by added manager 83
Guard Injection Guard the action by condition: checkout. qty > 0 + Order create . . . created qty : = checkout. qty. . . attr= oid, … Order(oid). qty>0 n Intuition: calculate and inject weakest precondition n GSM: guard-stage-milestone by IBM Nanjing U/2012 Summer School 2012/07/24 84
Conservative Injection ostat : = custsuppreply. ostat Order invokes custsupp further actiontaken action n If there is a shipment associated and is not finished v custsuppreply. ostat = CANCEL, violated v custsuppreply. ostat = CANCEL, consistent n Injection to further_action is FALSE Nanjing U/2012 Summer School 2012/07/24 85
Result n The injection is v Sound: strong enough to block potential violations v Conservative complete: weak enough to allow all possible updates that preserves the constraints in conservative manner checkout. qty = 0 attr checkout. qty = 10 checkout. custid = cust 001 Order + created ostat : = custsuppreply. ostat pay paid ostat : =CREAT; qty : = checkout. qty; custid : = checkout. custid; . . . Nanjing U/2012 Summer School 2012/07/24 further actiontaken action invokes custsupp 86
EZ-Flow and Research Problems verification dominance automated construction preserve data ICs runtime monitor dynamic modification process ICs exec. res. calculation Nanjing U/2012 Summer School 2012/07/24 87
Verification Problem n Given a biz process and a goal, do all executions of the workflow satisfy the goal? + Artifacts (Info models) + Semantic services (IOPEs) if C enable … ? = Conditionaction [Bhattacharya-Gerede-S. SOCA 07] [Gerede-S. ICSOC 07] [Bhattacharya-Gerede-Hull-Liu-S. BPM 07] [Deutsch-Hull-Patrizi-Vianu ICDT 09] [Vianu ICDT 09] Nanjing U/2012 Summer School 2012/07/24 88
Summary of Results n An artifact system W = ( G, S, R ) artifacts, services, rules n Ad hoc properties, restricted to defined-ness v Completion: Does W allow a complete run of an artifact? v Dead-end: Does W have a dead-end path? v Attribute redundancy: Does W have a redundant attribute? Undecidable in general, PSPACE if no artifact creation, intractable for monotonic workflows [Bhattacharya-Gerede-Hull-Liu-S. BPM 07] properties: LTL(FO) for guarded artifact schema v complete in PSPACE [Deutsch-Hull-Patrizi-Vianu ICDT 09] n Temporal Nanjing U/2012 Summer School 2012/07/24 89
EZ-Flow and Research Problems verification dominance automated construction preserve data ICs runtime monitor dynamic modification process ICs exec. res. calculation Nanjing U/2012 Summer School 2012/07/24 90
Outline n Challenges in Business Process Management n Artifact-centric n EZ-Flow Modeling Approach and Selected Technical Issues n Conclusions Nanjing U/2012 Summer School 2012/07/24 91
Conclusions n Biz process modeling: a foundation for many BPM issues v Many challenges: “old” and new v Data-centric or data aware approaches promising n Systematic exploration provides a good setting for the study v First step in a long march n Similar to my. SQL, will “my. BPM” be on the horizon? Nanjing U/2012 Summer School 2012/07/24 92
Acknowledgements n UCSB: Cagdas Gerede, Esra Kucukoguz, Yutian Sun n IBM: Rick Hull, Kamal Battacharya, Rong (Emily) Liu n Fudan U (China): Liang Zhang, Wei Xu, Yanguang Cheng, Haihuan Qin, Jiehui Li, Yi Lu n Christian Fritz (U Toronto-USC) n Diego Calvanese (U Bolzano) n Giuseppe De Giacomo (U Rome) n Jian Yang (Macquarie U) n Xi Liu (Nanjing U while visiting UCSB) Nanjing U/2012 Summer School 2012/07/24 93
Скачать презентацию Business Processes as Artifacts Jianwen Su University of
28bdfd733e9940789691a1cd791a3390.ppt