Download presentation
Presentation is loading. Please wait.
Published byColleen Wade Modified over 8 years ago
1
Workflows in Webdam Victor Vianu UC San Diego & INRIA/Webdam
2
Main results Specification and Verification PODS 2008, ICDT 2009, TODS 2009, ICDT 2011, BPM 2011, TODS 2012 Expressiveness ICDT 2011, TODS 2012 Collaborative workflows PODS 2013 Common thread: process + data ZOOM
3
Main results Specification and Verification PODS 2008, ICDT 2009, TODS 2009, ICDT 2011, BPM 2011, TODS 2012 Expressiveness ICDT 2011, TODS 2012 Collaborative workflows PODS 2013 Common thread: process + data
4
What is a data-centric workflow? States (data) Events (with data) Transitions state event
5
DB transition system....
6
Specifications of data-centric workflows Two formalisms used in work on verification: Active XML: XML with embedded functions documents evolve as a result of function calls Tuple Artifacts (IBM) relational database, evolving “artifact” tuple using pre/post conditions
7
What to verify: temporal properties of runs LTL + statements about data
8
LTL(FO) property: “every paid product is eventually delivered” G ( p F q) delivered (x) y (pay(x,y) price(x,y)) xx
9
Variant for AXML workflows G(p F q) X T p (X)T q (X) tree patterns with free variables X
10
Results on verification Tuple artifacts: V. + UCSD crowd Active XML: Serge + Luc + V. Restrictions that guarantee decidability
11
Main results Specification and verification PODS 2008, ICDT 2009, TODS 2009, ICDT 2011, BPM 2011, TODS 2012 Expressiveness ICDT 2011, TODS 2012 Collaborative workflows PODS 2013
12
Expressiveness: how to compare different workflows? Building a beehive Building an ant nest
13
Use Views! Building a beehiveBuilding an ant nest bee hive ant nest insect dwelling
14
Use Views! Building a beehiveBuilding an ant nest Map to a common abstraction Define a notion of simulation
15
Use Views! Building a beehiveBuilding an ant nest Allows to compare very different workflow formalisms
16
Example: AXML vs Tuple Artifacts AXML workflows can simulate Tuple Artifacts with respect to an appropriate view Tuple Artifacts cannot simulate AXML
17
Main results Specification and Verification PODS 2008, ICDT 2009, TODS 2009, ICDT 2011, BPM 2011, TODS 2012 Expressiveness ICDT 2011, TODS 2012 Collaborative workflows: PODS 2013
18
Collaborative workflows are pervasive Business processes, health care, scientific workflows, government …
19
tasks & honey actions & data AuthorPC chair PC member Referee Some possible actions Author: submit, respond, final, copyright PC chair: assign, discuss, decide, notify, remind PC member: enter review, invite referee, discuss Data-centric view: collaborative distributed data processing
20
virtual database local as view AuthorPC chair PC member Referee The collaborative workflow model Peer views projections of every relation including its key attributes tasks & honey actions & data
21
Peer actions conjunction of literals in the viewsequence of insertions/deletions into the view Update :- Condition fires with one valuation for variables PC Chair Assign(Id, PCmember), Status(Id, assigned) :- Submitted(Id, title, author, abstract), Status(Id, assigned) PC(PCmember), Conflict(author, PCmember)
22
action virtual database local as view AuthorPC chair PC member Referee tasks & honey actions & data
23
action virtual database local as view AuthorPC chair PC member Referee Inspired by Orchestra approach to data sharing [TJ Green et. al. SIGMOD 07] tasks & honey actions & data
24
Focus: runtime peer reasoning about global run based on local observations Monitoring events Diagnosing anomalous behavior Competitive advantage Has my paper been accepted? A paper was assigned to a PC member with a COI: could the assignment have been made by the PC chair, or was it necessarily made by a track chair? Is my own submission as a PC member still under discussion?
25
Focus: runtime peer reasoning about global run based on local observations Trace ρ Infinitely many compatible global runs
26
Specifying properties of global runs PLTL-FO PLTL: past linear-time propositional temporal logic PLTL-FO: PLTL with propositions interpreted as FO statements
27
Specifying properties of global runs PLTL-FO My paper (submission 12) has not yet been accepted nor rejected: G -1 review (Accept(12, review) Reject(12, review))
28
Specifying properties of global runs PLTL-FO My paper (submission 12) has not yet been accepted nor rejected : Semantics of a formula φ: relative to a trace ρ Cert(φ) : φ holds in every global run compatible with ρ Poss(φ) : φ holds in some global run compatible with ρ G -1 review (Accept(12, review) Reject(12, review))
29
Given a workflow specification W, a trace ρ and an PLTL-FO property φ of global runs evaluate Cert(φ) and Poss(φ) Goal Theorem: Cert(φ) and Poss(φ) can be evaluated in PSPACE with respect to φ and ρ.
30
Proof idea Trace ρ Infinitely many compatible global runs Symbolic transition system
31
Proof idea Trace ρ incomplete instances variant of C-tables variables + equality constraints
32
Proof idea Trace ρ Symbolic transitions among incomplete instances each computed in PSPACE
33
Proof idea Trace ρ Finite transition system on which Poss(φ) can be evaluated by a nondeterministic PSPACE search Cert(φ) = Poss( φ)
34
Other results Incremental evaluation Pre-emptive monitoring φ is not known a priori Can be done for classes of properties sharing a temporal type: underlying propositional PLTL formula Example: G -1 (p F -1 q) FO
35
Other results Incremental evaluation Pre-emptive monitoring Power of using knowledge about global run in workflow specifications PC member As long as I don’t know that my paper is accepted, I will reject every other paper Reject(ID) :- paper(ID, author), author ≠ self, cert (“my paper is accepted”) epistemic atom
36
Other results Incremental evaluation Pre-emptive monitoring Power of using knowledge about global run in workflow specifications Do epistemic atoms add power? Theorem: For our language, NO “epistemically closed”
37
Workflows in Webdam: Summary Models for data-centric workflows Verification and static analysis Framework for comparing expressiveness Collaborative workflows Better understanding of the impact of data and distribution in workflow modeling and analysis
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.