Architectural design: the coordination perspective

Transcription

1 Architectural design: the coordination perspective José Proença HASLab - INESC TEC & UM Arquitectura e Cálculo :FO

2 Outline 1.Visual semantics for Reo Connector colouring (CC)1 2.Locality (concurrency) partial connector colouring (PCC)2 3.Constraints SAT solving with data for Reo3 1 Dave Clarke, David Costa, and Farhad Arbab. Connector colouring I: Synchronisation and context dependency 2 Dave Clarke and José Proença. Partial connector colouring 3 Dave Clarke, José Proença, Alexander Lazovik, and Farhad Arbab, Channel-based coordination via constraint satisfaction José Proença, Dave Clarke, Interactive interaction constraints

3 9:35pm DST evening x y z Display x! bx := 9:35pm x $ y y! by := DST(bx) (y ^ evening(by)) $ z z! bz := by Reo as (Interactive) Constraints

4 Coordination as constraint satisfaction a b a $ b b! b b := ba Formulas a b b! a b! b b := ba update solve a P b b! b b := ba (a ^ P(ba)) $ b Solution a f b a $ b b! b b := f(ba)

5 Exercise: write constraints Wr(d) Rd

6 Building up constraints Connector colouring as constraints Data constraints Interactive constraints

7 Context (need for extra interaction)

8 Hotel booking Request SrchHotel 1 SrchHotel 2 Approve Book [Cancel] Confirmed SrchHotel 3 Interaction with hotel repositories Interaction with users Interaction with hotels (availability & payment)

9 Problem of Interaction

10 Coordination as constraint satisfaction Constraint update [interaction] solve Solution 0 dataflow 0

11 Asynchronous transactions in A i F i ok i out ok i stopped F 1 i stop in stopped A 1 A 2 A n out stop

12 Synchronous transactions in S i F i ok i out stopped F 1 i ok i stop in stopped S 1 S 2 S n out stop

13 Problem Interaction or Synchrony but not both

14 Solution Interactive Interaction Constraints

15 Coordination as constraint satisfaction Constraint update [interaction] solve Solution dataflow

16 Coordination as interactive constraint satisfaction Constraint update [interaction] solve [interaction] Solution dataflow

17 Synchronous transactions in S i F i ok i out stopped F 1 i ok i stop in stopped S 1 S 2 S n out stop

18 Synchronous transactions in B i F i [F 1 i ] ok i out in B 1 B 2 B n out

19 Interactive Interaction Constraints DB CA User Wr Wr Rd

20 Scala / Java implementation in B i F i [F 1 i ] ok i out val f = Function( f ) { case s: String => /* do something */ } val finv = Function( f^-1 ) { case s: String => /* do something */ } class Filter(a: String, b: String, p: Predicate) extends... { val ok = Predicate( ok ) { case s: String => /* do something */ } val connector = writer( in,list( a, b )) ++ transf( in, x,f,finv) ++ filter( x, out,ok) ++ sdrain( x, out ) ++ reader( out,2) connector.run() } def getconstraint = Constraint( b --> a, b --> (b := a), b --> (a :< p), (a /\ (a :< p)) --> b ) /* override def update(s: Option[Solution]) =... */

21 More about constraints

22 CC2 as constraints [ Colouring: End {Flow, NoFlow} ] Formula: Boolean over End Composition = conjunction a b ( a b) a b (a b) b! a a b (a b)

23 CC3 as constraints [ Colouring: End {Flow, GiveReason, GetReason} ] Formula: Boolean over End, End src, Endsnk a = flow on a ; a src = give reason ; bsnk = get reason a a a a b b b b b a (a b) (a b snk ) a ( b a src b snk )

24 CC3 as constraints Composition a = flow on a ; asrc = give reason ; bsnk = get reason one-to-one composition: source to sink ends x x snk x src

25 CC3 as constraints example a b c = b a c (a b) (a b snk ) a ( b a src b snk ) ( b b src ) c snk b src b snk {a b c c snk } = { a b c a src b snk b src c snk } =

26 Data constraints 9:35pm DST evening x y z Display x! bx := 9:35pm x $ y y! by := DST(bx) (y ^ evening(by)) $ z z! bz := by How to solve this?

27 Predicate abstraction 9:35pm DST evening x y z Display original x! bx := 9:35pm x $ y y! by := DST(bx) (y ^ evening(by)) $ z z! bz := by boolean x! bx ev.dst :=[evening(dst(9:35pm))] y! by ev := bx ev.dst (y ^ by ev ) $ z x $ y

28 Interaction via Choco 9:35pm DST evening x y z Display boolean x! bx ev.dst :=[evening(dst(9:35pm))] y! by ev := bx ev.dst (y ^ by ev ) $ z x $ y interactive x! XPred(ev.dst,x,9:35pm) x $ y y! by ev := bx ev.dst (y ^ by ev ) $ z

29 Interaction via Choco 9:35pm boolean x y z Display x! bx ev.dst :=[evening(dst(9:35pm))] interactive instance of a Choco constraint DST evening reacts when x or bx ev.dst is instantiated x bx can be anything ev.dst x ) bx ev.dst = ev(dst(9:35pm)) y! by ev := bx ev.dst x! XPred(ev.dst,x,9:35pm) y! by ev := bx ev.dst (y ^ by ev ) $ z x $ y (y ^ by ev ) $ z x $ y

30 Wrapping up Rd Reo and connector colouring Interactive constraint solving in stopped A i F i F 1 i ok i ok i out stop Expose the atomicity of Reo to components in B i F i [F 1 i ] ok i out

31 Ongoing experiments Avoiding pre-processing (SMT instead of SAT) a, â Compiling steps: Heuristics for identifying potential partial colourings Combining local and interactive constraints (probably to a journal)