Distributed Synthesis for Well Connected Architectures Paul Gastin, - - PowerPoint PPT Presentation

distributed synthesis for well connected architectures
SMART_READER_LITE
LIVE PREVIEW

Distributed Synthesis for Well Connected Architectures Paul Gastin, - - PowerPoint PPT Presentation

Distributed Synthesis for Well Connected Architectures Paul Gastin, Nathalie Sznajder and Marc Zeitoun March 13th 2006 ACI Cortos Persee Versydis Synthesis of a reactive system inputs from E outputs to E Specification Open system S


slide-1
SLIDE 1

Distributed Synthesis for Well Connected Architectures

Paul Gastin, Nathalie Sznajder and Marc Zeitoun March 13th 2006 ACI Cortos Persee Versydis

slide-2
SLIDE 2

Synthesis of a reactive system

inputs from E

  • utputs to E

Open system S Specification ϕ

slide-3
SLIDE 3

Synthesis of a reactive system

inputs from E

  • utputs to E

Open system S Specification ϕ Program P

Two problems

Decide whether there exists a program st. P||E | = ϕ, ∀E. Synthesis: If so, compute such a program. For reasonable systems and specifications, the problems are decidable.

slide-4
SLIDE 4

Distributed synthesis

input of E

  • utput to E

Open distributed system S S1 S2 S3 S4 Specification ϕ

slide-5
SLIDE 5

Distributed synthesis

input of E

  • utput to E

Open distributed system S S1 S2 S3 S4 Specification ϕ P1 P2 P3 P4

Two problems

Decide the existence of a distributed program such that their joint behavior P1||P2||P3||P4||E satisfies ϕ, for all E. Synthesis : If it exists, compute such a distributed program.

slide-6
SLIDE 6

Distributed synthesis

input of E

  • utput to E

Open distributed system S S1 S2 S3 S4 Specification ϕ P1 P2 P3 P4

Two problems

Decide the existence of a distributed program such that their joint behavior P1||P2||P3||P4||E satisfies ϕ, for all E. Synthesis : If it exists, compute such a distributed program.

Peterson-Reif 1979, Pnueli-Rosner 1990

In general, the problem is undecidable.

slide-7
SLIDE 7

The model

Example

x0 x1 x2 x3 x4 x5 a1 a2 a3 a4

slide-8
SLIDE 8

The model

Example

x0 x1 x2 x3 x4 x5 a1 a2 a3 a4 Synchronous behavior

slide-9
SLIDE 9

The model

Example

x0 x1 x2 x3 x4 x5 a1 a2 a3 a4 Synchronous behavior Strategies with local memory

slide-10
SLIDE 10

The model

Example

x0 x1 x2 x3 x4 x5 a1 a2 a3 a4 Synchronous behavior Strategies with local memory 0-delay semantics

slide-11
SLIDE 11

The model

Example

x0 x1 x2 x3 x4 x5 a1 a2 a3 a4 Synchronous behavior Strategies with local memory 0-delay semantics Input-output specifications

slide-12
SLIDE 12

Undecidable architecture (Pnueli–Rosner ’90)

x0 y0 x1 y1 P0 P1

slide-13
SLIDE 13

Undecidable architecture (Pnueli–Rosner ’90)

x0 y0 x1 y1 P0 P1 S

slide-14
SLIDE 14

Undecidable architecture (Pnueli–Rosner ’90)

x0 y0 x1 y1 P0 P1 S $Ci$

T

$C ′

i $

slide-15
SLIDE 15

Undecidable architecture (Pnueli–Rosner ’90)

x0 y0 x1 y1 P0 P1 S $Ci$

T

$C ′

i $

$Ci+1$ ⇒ T $C ′

i+1$

slide-16
SLIDE 16

Decidable architecture

x0 y0 x1 y1 t P0 P1

slide-17
SLIDE 17

Pipe-line decidable for global specifications (Kupferman–Vardi ’01)

Oenv P1 P2 . . . Pn On O1 O2 On−1

slide-18
SLIDE 18

Pipe-line decidable for global specifications (Kupferman–Vardi ’01)

Oenv P1 P2 . . . Pn On O1 O2 On−1

slide-19
SLIDE 19

Pipe-line decidable for global specifications (Kupferman–Vardi ’01)

Oenv P1 P2 . . . Pn On O1 O2 On−1

slide-20
SLIDE 20

Information fork criterion (Finkbeiner–Schewe ’05)

x0 x1 p0 p1 t0 t1 p y

slide-21
SLIDE 21

Information fork criterion (Finkbeiner–Schewe ’05)

x0 x1 p0 p1 t0 t1 p y x0 x1 p0 p1 t0 t1

slide-22
SLIDE 22

Outline

1

Uncomparable information

2

Uniformly well connected architectures

3

Well connected architectures

slide-23
SLIDE 23

Outline

1

Uncomparable information

2

Uniformly well connected architectures

3

Well connected architectures

slide-24
SLIDE 24

Uncomparable information yields undecidability

Definition

For an output variable v, View(v) is the set of input variables u such that v is accessible from u.

Definition

An architecture has uncomparable information if there exist x,y output variables such that View(x) \ View(y) = ∅ and View(y) \ View(x) = ∅. Otherwise it is said to have preordered information. x0 x1 y0 y1

slide-25
SLIDE 25

Uncomparable information yields undecidability

Theorem

Architectures with uncomparable information are undecidable for LTL or CTL input-output specifications.

Proof

LTL specifications : x0 x1 y0 y1 x0 x1 y0 y1

slide-26
SLIDE 26

Uncomparable information yields undecidability

Theorem

Architectures with uncomparable information are undecidable for LTL or CTL input-output specifications.

Proof

LTL specifications : x0 x1 y0 y1 x0 x1 y0 y1

slide-27
SLIDE 27

Uncomparable information yields undecidability

Theorem

Architectures with uncomparable information are undecidable for LTL or CTL input-output specifications.

Proof

LTL specifications : x0 x1 y0 y1 x0 x1 y0 y1 0 0 0 0 0

slide-28
SLIDE 28

Outline

1

Uncomparable information

2

Uniformly well connected architectures

3

Well connected architectures

slide-29
SLIDE 29

Uniformly well connected architectures

Definition

An architecture is uniformly well connected if there is a uniform way to route variables in View(v) to v for each output variable v. u v w p p s t p p p x y z

slide-30
SLIDE 30

Uniformly well connected architectures

Definition

An architecture is uniformly well connected if there is a uniform way to route variables in View(v) to v for each output variable v. u v w p p s t p p p x y z u ⊕ v v ⊕ w

slide-31
SLIDE 31

Network Information Flow

M M M M s s s a b c d e f g

demands M, M, M, M demands M, M

slide-32
SLIDE 32

Network Information Flow : Multicast

M M M M s a b c d e f g

demands M, M, M, M demands M, M, M, M

slide-33
SLIDE 33

Relations between the two problems

Theorem

The multicast reduces to checking uniform well connectedness.

slide-34
SLIDE 34

Relations between the two problems

Theorem

The multicast reduces to checking uniform well connectedness.

Proof.

M M

s a b c d e f

slide-35
SLIDE 35

Relations between the two problems

Theorem

The multicast reduces to checking uniform well connectedness.

Proof.

M M

s a b c d e f

M M

s a b c d e f u v w x y z s t r y y y

slide-36
SLIDE 36

Relations between the two problems

Theorem

Checking uniform well connectedness reduces to the network information flow.

slide-37
SLIDE 37

Relations between the two problems

Theorem

Checking uniform well connectedness reduces to the network information flow.

Proof.

s

slide-38
SLIDE 38

Relations between the two problems

Theorem

Checking uniform well connectedness reduces to the network information flow.

Proof.

s s

demands s demands s

slide-39
SLIDE 39

Relations between the two problems

Theorem

Checking uniform well connectedness reduces to the network information flow.

Proof.

s s

demands s demands s

slide-40
SLIDE 40

Relations between the two problems

Theorem

Checking uniform well connectedness reduces to the network information flow.

Proof.

s s

demands s demands s

slide-41
SLIDE 41

Complexity

Rasala Lehman-Lehman 2004

Multicast whith alphabet size q = pk (where p is prime) is NP-hard.

slide-42
SLIDE 42

Complexity

Rasala Lehman-Lehman 2004

Multicast whith alphabet size q = pk (where p is prime) is NP-hard.

Theorem

Checking whether a given architecture is uniformly well connected is NP-complete.

slide-43
SLIDE 43

Complexity

Rasala Lehman-Lehman 2004

Multicast whith alphabet size q = pk (where p is prime) is NP-hard.

Theorem

Checking whether a given architecture is uniformly well connected is NP-complete.

Proof.

It is trivially NP. Reduction from multicast gives NP-hardness.

slide-44
SLIDE 44

Decidability criterion for uniformly well connected architectures

Theorem

Architectures with preordered information are decidable for CTL* specifications.

slide-45
SLIDE 45

Decidability criterion for uniformly well connected architectures

Theorem

Architectures with preordered information are decidable for CTL* specifications.

Proof.

x1 y1 x2 y2 x3 y3 x4 y4

slide-46
SLIDE 46

Decidability criterion for uniformly well connected architectures

Theorem

Architectures with preordered information are decidable for CTL* specifications.

Proof.

x1 y1 x2 y2 x3 y3 x4 y4

slide-47
SLIDE 47

Decidability criterion for uniformly well connected architectures

Theorem

Architectures with preordered information are decidable for CTL* specifications.

Proof.

x1 y1 x2 y2 x3 y3 x4 y4

slide-48
SLIDE 48

Decidability criterion for uniformly well connected architectures

Theorem

Architectures with preordered information are decidable for CTL* specifications.

Proof.

x1 y1 x2 y2 x3 y3 x4 y4

slide-49
SLIDE 49

Decidability criterion for uniformly well connected architectures

Theorem

Architectures with preordered information are decidable for CTL* specifications.

Proof.

x1 y1 x2 y2 x3 y3 x4 y4 y1 y2 y3 y4 a1 a2 a3 a4 x1 x2 x3 x4 x2 x3 x4 x3 x4 x4

Theorem: Kupferman-Vardi (LICS’01)

The synthesis problem with local strategies is decidable for pipeline archi- tectures and CTL∗ specifications (or tree-automata specifications) on all variables.

slide-50
SLIDE 50

Outline

1

Uncomparable information

2

Uniformly well connected architectures

3

Well connected architectures

slide-51
SLIDE 51

Well connected architectures

Definition

An architecture is well connected if, for each output variable y, the subarchitecture formed by E ∗−1(y) is uniformly well connected. u v w p p s t p p p x y z

slide-52
SLIDE 52

Well connected architectures

Definition

An architecture is well connected if, for each output variable y, the subarchitecture formed by E ∗−1(y) is uniformly well connected. u v w p p s t p p p x y z v

slide-53
SLIDE 53

Well connected architectures

Definition

An architecture is well connected if, for each output variable y, the subarchitecture formed by E ∗−1(y) is uniformly well connected. u v w p p s t p p p x y z u w

slide-54
SLIDE 54

Well connected architectures

Definition

An architecture is well connected if, for each output variable y, the subarchitecture formed by E ∗−1(y) is uniformly well connected. u v w p p s t p p p x y z v

slide-55
SLIDE 55

Well connected architectures

Theorem

One can decide whether an architecture is well-connected in polynomial time.

slide-56
SLIDE 56

Well connected architectures

Theorem

One can decide whether an architecture is well-connected in polynomial time.

slide-57
SLIDE 57

Well connected architectures

Theorem

One can decide whether an architecture is well-connected in polynomial time.

slide-58
SLIDE 58

Well connected architectures

Theorem

One can decide whether an architecture is well-connected in polynomial time.

Rasala Lehman–Lehman 2004

One can solve the network information flow in the special case where there is a unique sink in polynomial time.

slide-59
SLIDE 59

Well connected architectures

Theorem

There exists a well connected architecture which is not uniformly well connected.

slide-60
SLIDE 60

Well connected architectures

Theorem

There exists a well connected architecture which is not uniformly well connected.

Lemma (Rasala Lehman–Lehman 2004)

If f 1, . . . , f n are pairwise independent functions of the form S2 → S, then n ≤ |S| + 1.

slide-61
SLIDE 61

Well connected architectures

Theorem

There exists a well connected architecture which is not uniformly well connected.

Lemma (Rasala Lehman–Lehman 2004)

If f 1, . . . , f n are pairwise independent functions of the form S2 → S, then n ≤ |S| + 1. u w z1 z2 z3 z4 z12 z13 z14 z23 z24 z34

slide-62
SLIDE 62

Well connected preordered architectures are undecidable

Theorem

The synthesis problem for LTL specifications and well connected, preordered information is undecidable.

slide-63
SLIDE 63

Well connected preordered architectures are undecidable

Theorem

The synthesis problem for LTL specifications and well connected, preordered information is undecidable. w u v x y z0 q0 p6 p z1 z2 z3 z4 p1 p2 p3 p4 p5 u1 w1 u2 w2 u3 w3 u4 w4 u5 w5 u6 w6

slide-64
SLIDE 64

Well connected preordered architectures are undecidable

Theorem

The synthesis problem for LTL specifications and well connected, preordered information is undecidable. w u v x y z0 q0 p6 p z1 z2 z3 z4 p1 p2 p3 p4 p5 u1 w1 u2 w2 u3 w3 u4 w4 u5 w5 u6 w6

slide-65
SLIDE 65

Well connected preordered architectures are undecidable

Theorem

The synthesis problem for LTL specifications and well connected, preordered information is undecidable. w u v x y z0 q0 p6 p z1 z2 z3 z4 p1 p2 p3 p4 p5 u1 w1 u2 w2 u3 w3 u4 w4 u5 w5 u6 w6 0q1p0{0, 1}ω #p+qCp#ω

slide-66
SLIDE 66

Well connected preordered architectures are undecidable : The specification

w u v x y z0 p0 p6 p z1 z2 z3 z4 p1 p2 p3 p4 p5 u1 w1 u2 w2 u3 w3 u4 w4 u5 w5 u6 w6

slide-67
SLIDE 67

Well connected preordered architectures are undecidable : The specification

w u v x y z0 p0 p6 p z1 z2 z3 z4 p1 p2 p3 p4 p5 u1 w1 u2 w2 u3 w3 u4 w4 u5 w5 u6 w6 0q10{0, 1}ω #q+1C1#ω 0q′10{0, 1}ω #q′+1C1#ω

slide-68
SLIDE 68

Well connected preordered architectures are undecidable : The specification

w u v x y z0 p0 p6 p z1 z2 z3 z4 p1 p2 p3 p4 p5 u1 w1 u2 w2 u3 w3 u4 w4 u5 w5 u6 w6 0q1p0{0, 1}ω #q+pC#ω 0q′1p′0{0, 1}ω #q′+p′C′#ω

slide-69
SLIDE 69

Well connected preordered architectures are undecidable : The specification

w u v x y z0 p0 p6 p z1 z2 z3 z4 p1 p2 p3 p4 p5 u1 w1 u2 w2 u3 w3 u4 w4 u5 w5 u6 w6 0q1p0{0, 1}ω 0q1p0{0, 1}ω #q+pC#ω #q+pC′#ω u = v = ⇒ x = y

slide-70
SLIDE 70

Well connected preordered architectures are undecidable : The specification

w u v x y z0 p0 p6 p z1 z2 z3 z4 p1 p2 p3 p4 p5 u1 w1 u2 w2 u3 w3 u4 w4 u5 w5 u6 w6 0q1p+10{0, 1}ω 0q+11p0{0, 1}ω #q+p+1C#ω #q+p+1C′#ω u = v + 1 = ⇒ C′ ⊢ C

slide-71
SLIDE 71

Well connected preordered architectures are undecidable : The specification

w u v x y z0 p0 p6 p z1 z2 z3 z4 p1 p2 p3 p4 p5 u1 w1 u2 w2 u3 w3 u4 w4 u5 w5 u6 w6 0 . . . 01 u w u w u w u w u w w

slide-72
SLIDE 72

Well connected preordered architectures are undecidable : A distributed implementation

w u v x y z0 p0 p6 p z1 z2 z3 z4 p1 p2 p3 p4 p5 u1 w1 u2 w2 u3 w3 u4 w4 u5 w5 u6 w6

slide-73
SLIDE 73

Well connected preordered architectures are undecidable : A distributed implementation

w u v x y z0 p0 p6 p p Cp q Cq z1 z2 z3 z4 p1 p2 p3 p4 p5 u1 w1 u2 w2 u3 w3 u4 w4 u5 w5 u6 w6

slide-74
SLIDE 74

Well connected preordered architectures are undecidable : A distributed implementation

w u v x y z0 p0 p6 p p Cp q Cq 0 . . . 01 u u ⊕ w u w w u w u w u w u w u w w z1 z2 z3 z4 p1 p2 p3 p4 p5 u1 w1 u2 w2 u3 w3 u4 w4 u5 w5 u6 w6

slide-75
SLIDE 75

Well connected preordered architectures are undecidable : Another distributed implementation

w u v x y z0 p0 p6 p p Cp q Cq 0 . . . 01 u u u ⊕ w w Y u ⊕ w u w u w u w u w u w w z1 z2 z3 z4 p1 p2 p3 p4 p5 u1 w1 u2 w2 u3 w3 u4 w4 u5 w5 u6 w6

slide-76
SLIDE 76

Well connected preordered architectures are undecidable : The specification (end)

w u v x y z0 p0 p6 p z1 z2 z3 z4 p1 p2 p3 p4 p5 u1 w1 u2 w2 u3 w3 u4 w4 u5 w5 u6 w6 0 . . . 01 . . . u w

slide-77
SLIDE 77

Well connected preordered architectures are undecidable: Masking one bit of u to y

Lemma

Let ub = 0qb1p0u′ and w = 0q1w′. Then (ˆ fz3, ˆ fz4)(u0[n], w[n]) = (ˆ fz3, ˆ fz4)(u1[n], w[n]) w u v x y z0 p0 p6 p z1 z2 z3 z4 p1 p2 p3 p4 p5 u1 w1 u2 w2 u3 w3 u4 w4 u5 w5 u6 w6

slide-78
SLIDE 78

Well connected preordered architectures are undecidable: Masking one bit of u to y

Lemma

Let ub = 0qb1p0u′ and w = 0q1w′. Then (ˆ fz3, ˆ fz4)(u0[n], w[n]) = (ˆ fz3, ˆ fz4)(u1[n], w[n]) w 0q1{0, 1}ω u p ∼ 0q01p0{0, 1}ω p + 1 ∼ 0q11p0{0, 1}ω v x y z0 p0 p6 p z1 z2 z3 z4 p1 p2 p3 p4 p5 u1 w1 u2 w2 u3 w3 u4 w4 u5 w5 u6 w6

slide-79
SLIDE 79

Well connected preordered architecures are undecidable: Enforcing output of the correct configuration

Lemma

For all q ≥ 0, u ∈ 0q1p0{0, 1}ω = ⇒ x = #q+pCp#ω w u v x y z0 p0 p6 p z1 z2 z3 z4 p1 p2 p3 p4 p5 u1 w1 u2 w2 u3 w3 u4 w4 u5 w5 u6 w6

slide-80
SLIDE 80

Well connected preordered architecures are undecidable: Enforcing output of the correct configuration

Lemma

For all q ≥ 0, u ∈ 0q1p0{0, 1}ω = ⇒ x = #q+pCp#ω w u p + 1 ∼ 0q11p0{0, 1}ω v x y z0 p0 p6 p z1 z2 z3 z4 p1 p2 p3 p4 p5 u1 w1 u2 w2 u3 w3 u4 w4 u5 w5 u6 w6

slide-81
SLIDE 81

Well connected preordered architecures are undecidable: Enforcing output of the correct configuration

Lemma

For all q ≥ 0, u ∈ 0q1p0{0, 1}ω = ⇒ x = #q+pCp#ω w 0q1{0, 1}ω u p + 1 ∼ 0q11p0{0, 1}ω v 0q+11p0{0, 1}ω x y z0 p0 p6 p z1 z2 z3 z4 p1 p2 p3 p4 p5 u1 w1 u2 w2 u3 w3 u4 w4 u5 w5 u6 w6

slide-82
SLIDE 82

Well connected preordered architecures are undecidable: Enforcing output of the correct configuration

Lemma

For all q ≥ 0, u ∈ 0q1p0{0, 1}ω = ⇒ x = #q+pCp#ω w 0q1{0, 1}ω u p + 1 ∼ 0q11p0{0, 1}ω v 0q+11p0{0, 1}ω p ∼ 0q01p0{0, 1}ω 0q+11p0{0, 1}ω 0q1{0, 1}ω x y z0 p0 p6 p z1 z2 z3 z4 p1 p2 p3 p4 p5 u1 w1 u2 w2 u3 w3 u4 w4 u5 w5 u6 w6

slide-83
SLIDE 83

Well connected preordered architecures are undecidable: Enforcing output of the correct configuration

Lemma

For all q ≥ 0, u ∈ 0q1p0{0, 1}ω = ⇒ x = #q+pCp#ω w 0q1{0, 1}ω u p + 1 ∼ 0q11p0{0, 1}ω v 0q+11p0{0, 1}ω p ∼ 0q01p0{0, 1}ω 0q+11p0{0, 1}ω 0q1{0, 1}ω x #p+q+1Cp#ω y z0 p0 p6 p z1 z2 z3 z4 p1 p2 p3 p4 p5 u1 w1 u2 w2 u3 w3 u4 w4 u5 w5 u6 w6

slide-84
SLIDE 84

Well connected preordered architecures are undecidable: Enforcing output of the correct configuration

Lemma

For all q ≥ 0, u ∈ 0q1p0{0, 1}ω = ⇒ x = #q+pCp#ω w 0q1{0, 1}ω u p + 1 ∼ 0q11p0{0, 1}ω v 0q+11p0{0, 1}ω p ∼ 0q01p0{0, 1}ω 0q+11p0{0, 1}ω 0q1{0, 1}ω x #p+q+1Cp#ω #p+q+1Cp#ω y z0 p0 p6 p z1 z2 z3 z4 p1 p2 p3 p4 p5 u1 w1 u2 w2 u3 w3 u4 w4 u5 w5 u6 w6

slide-85
SLIDE 85

Well connected preordered architecures are undecidable: Enforcing output of the correct configuration

Lemma

For all q ≥ 0, u ∈ 0q1p0{0, 1}ω = ⇒ x = #q+pCp#ω w 0q1{0, 1}ω u p + 1 ∼ 0q11p0{0, 1}ω v 0q+11p0{0, 1}ω p ∼ 0q01p0{0, 1}ω 0q+11p0{0, 1}ω 0q1{0, 1}ω x #p+q+1Cp#ω #p+q+1Cp#ω #p+q+1Cp#ω y z0 p0 p6 p z1 z2 z3 z4 p1 p2 p3 p4 p5 u1 w1 u2 w2 u3 w3 u4 w4 u5 w5 u6 w6

slide-86
SLIDE 86

Well connected preordered architecures are undecidable: Enforcing output of the correct configuration

Lemma

For all q ≥ 0, u ∈ 0q1p0{0, 1}ω = ⇒ x = #q+pCp#ω w 0q1{0, 1}ω u p + 1 ∼ 0q11p0{0, 1}ω v 0q+11p0{0, 1}ω p ∼ 0q01p0{0, 1}ω 0q+11p0{0, 1}ω 0q1{0, 1}ω x #p+q+1Cp#ω #p+q+1Cp#ω #p+q+1Cp#ω #p+q+1Cp+1#ω y z0 p0 p6 p z1 z2 z3 z4 p1 p2 p3 p4 p5 u1 w1 u2 w2 u3 w3 u4 w4 u5 w5 u6 w6

slide-87
SLIDE 87

April Rasala Lehman and Eric Lehman. Complexity classification of network information flow problems. In SODA, pages 142–150, 2004.