Synchronization strategies for global computing

1
Synchronization strategiesfor global computing
Thesis Progress Report
Ivan Lanese Computer Science Department University
of Pisa
Relator
Prof. Ugo Montanari
2
Roadmap

• Global computing
• SHR and other formal tools
• What we did in the first year
• What we planned after the first year
• What we have done
• What is still to be done

3
Roadmap

• Global computing
• SHR and other formal tools
• What we did in the first year
• What we planned after the first year
• What we have done
• What is still to be done

4
What is global computing?

• Essentially networks deployed on huge areas
• It is becoming the most widespread computing
  paradigm
• Internet, wireless communication networks,
  overlay networks

5
Features of global computing systems

• Distribution
• Huge areas localities can not always be hidden
• Mobility
• Both physical and code mobility
• Eterogeneity
• Interoperability, coordination
• Openness
• Reconfigurability

6
Formal methods for GC

• Building models of the system
• To concentrate on a particular aspect
• To abstract from details
• To analyze the properties of the system before
  building it
• Traditional formal methods are not enough for GC
• Mobility must be modeled explicitly
• Need for compositionality
• Need for more abstraction

7
High level models

• We look for models at high level of abstraction
• Models of coordination among subsystems
• For making the model of the whole system
  treatable
• We need powerful primitives
• Multiple synchronizations
• Abstractions of full protocols
• Declarative specification of constraints
• Possible evolutions derived as solution of system
  of constraints

8
Roadmap

• Global computing
• SHR and other formal tools
• What we did in the first year
• What we planned after the first year
• What we have done
• What is still to be done

9
Synchronized Hyperedge Replacement

• Follows the traditional approach of graph
  transformation
• Deals well with distribution, mobility,
  compositionality
• (Hyper)edges are systems connected through common
  nodes
• Productions describe the evolution of single
  edges
• Local effect, easy to implement
• Synchronization via constraints on nodes
• Global constraint solving algorithm to find
  allowed transitions
• Productions applied indipendently
• Allows to define complex transformations
• Allows multiple synchronization
• Declarative approach

10
Hyperedge Replacement Systems

• A production describes how the hyperedge L is
  rewritten into the graph R

L
R
H
3
3
4
4
2
2
1
1
11
Hyperedge Replacement Systems

• A production describes how the hyperedge L is
  transformed into the graph R

Many concurrent rewritings are allowed
12
Synchronizing productions

• Synchronized transitions we associate actions
  to nodes. A transition is allowed iff the
  synchronization constraints imposed on actions
  are satisfied
• Many synchronization models are possible (Hoare,
  Milner, ...)

13
An example Milner SHR

• Milner synchronization pair of edges can
  synchronize by doing complementary actions

14
SHR with mobility
15
Other formal tools CommUnity in one slide

• Architectural model of concurrent systems
• Based on communication via channels and
  synchronization of actions
• Separation between computation and coordination
  concerns
• Computation inside programs
• Coordination explicitly represented via morphisms
  and special programs (cables)
• Composition via colimits

16
Other formal tools Tile Model in one slide

• Extension of SOS systems and rewriting logic that
  allows synchronization
• Rules have trigger and effect
• Operators of horizontal, vertical and monoidal
  composition
• Useful for studying observational properties of
  open systems

17
Roadmap
old

• Global computing
• SHR and other formal tools
• What we did in the first year
• What we planned after the first year
• What we have done
• What is still to be done

18
Comparing models for GC
old

• Comparing synchronization strategies, mobility
  strategies and binding operators in different
  models
• Comparative analysis of the following paradigms
• Fusion Calculus calculus for mobility, allows
  fusions
• SHR already presented in detail
• Logic programming developed programming
  paradigm, we use it as goal rewriting mechanism,
  to be extended to deal with GC systems
• Mapping of Fusion Calculus into SHR and of SHR
  into logic programming
• Fusion Calculus is a subset of Milner SHR
• Logic programming (with transactions) related to
  Hoare SHR

19
Comparing strategies for compositionality
old

• Two approaches to compositionality
• Categorical approach via universal constructions
  (colimits)
• Algebraic approach via operators for building
  systems
• Our case study CommUnity vs Tile Model

20
Results from CommUnity vs tiles
old

• Semantics for CommUnity
• Decomposition of CommUnity programs
• New algebra of connectors (in the Tile Model)
  that simulates CommUnity synchronization
• Correlation between colimit and algebraic aspects
  (bisimulation, axiomatization)

21
Roadmap
plan

• Global computing
• SHR and other formal tools
• What we did in the first year
• What we planned after the first year
• What we have done
• What is still to be done

22
A physical analogy
plan

• Physical systems
• Many interacting components
• Partial derivative equations for parts of the
  space
• Components of the system gives boundary
  conditions
• The solutions of the system of equations are the
  allowed behaviors
• Global computing systems
• Equations associated to the interfaces
• Equations are a declarative way of specifying
  coordination
• Only simple equations used in past works
• Hoare and Milner synchronization in SHR
• Developing new models using this analogy

23
Different items
plan

• Extending synchronization algebras with mobility
• Standard synchronization algebras allow to
  categorize synchronization models
• Need extensions to be applied to GC models
• SHR well-suited for these analysis
• Analyzing observational semantics of GC models
• Allows to analyze incomplete systems
  (compositionality)
• Tile Logic well-suited for such analysis
• Hybrid models
• For taking the best features of different models
• Suggested by the comparisons done

24
General aim
plan

• Double aim
• Analyzing existing models and their relationships
• Looking for new models for GC systems

25
Roadmap
ok

• Global computing
• SHR and other formal tools
• What we did in the first year
• What we planned after the first year
• What we have done
• What is still to be done

26
SAMs (1)
ok

• Synchronization algebras with mobility
• Extend synchronization algebras to deal with name
  mobility and local resources
• Different synchronization policies allowed
• Simple ones Milner, Hoare, broadcast
• More complex ones with priority, treshold
  synchronization
• Like differential equations for physics

27
SAMs (2)
ok

• SAs specify composition of actions
• SAMs also gives
• Mapping from parameters of synchronizing actions
  and parameters of the result
• Fusions among parameters
• Final actions (performed on local channels)
• Some more technical stuff

28
An example SAM for Milner synchronization
ok

• Actions normal actions, coactions, tau, epsilon
• Normal actions synchronize with corresponding
  coactions giving tau, corresponding parameters
  are fused, no parameters are propagated
• Anything synchronize with epsilon, action and
  parameters are propagated, no fusions
• No other synchronization is allowed
• Only tau and epsilon can be done on local channels

29
Applying SAMs
ok

• Applied in the context of SHR
• Parametric SHR
• The SAM is a parameter of the model
• Different models obtained via instantiation
• Properties can be proved in general
• Heterogeneous SHR
• Each node is labelled by a SAM
• SAMs are required to form a commutative monoid
• Node fusions cause SAMs composition
• Broadcast fusion calculus
• Via the mapping discovered in the first year

30
Abstract semantics for SHR
ok

• Bisimulation for (parametric) SHR
• Congruence results w.r.t. a suitable algebra of
  graphs
• Found using the Tile Model as framework, but then
  provable indipendently

31
Extending CommUnity vs Tile Model (1)
ok

• Analysis of the class of connectors for CommUnity
• 3 different semantics for connectors
• observational
• axiomatic
• denotational
• The 3 semantics are equivalent
• Study of the expressivity of different classes of
  connectors
• The larger class allows to produce a connector
  for each denotation

? 0 1
00 tick untick
01 untick untick
10 untick untick
11 untick tick
32
Extending CommUnity vs Tile Model (2)
ok

• Double correspondance between categorical and
  algebraic approach
• Starting diagram and colimit equals up to
  bisimulation
• Axiomatization of bisimulation and reduction to
  normal form allow to compute colimit algebraically

33
Roadmap
todo

• Global computing
• SHR and other formal tools
• What we did in the first year
• What we planned after the first year
• What we have done
• What is still to be done

34
From planned work
todo

• More refined abstract semantics
• Abstract semantics for SHR is very (too)
  distinguishing
• Hybrid models
• Models with different synchronization policies
  have been derived but their properties have to be
  analyzed in more depth
• What about other hybrid models?
• SAMs
• SAMs have been applied to SHR
• Can them be applied to other formalisms
  (process-calculi)?

35
New ideas
todo

• Refining
• To bridge the gap between high level models and
  implementation
• Must preserve behavioral properties
• Combining SHR and bigraphs
• Bigraphs and SHR have the same field of
  application but different aim (double structure
  vs compositionality)

36
Bibliography (old)

• Software Architecture, Global Computing and
  Graph Transformation via Horn Clauses, I. Lanese
  and U. Montanari, Proceedings SBES 2002 16th
  Brazilian Symposium on Software Engineering
• A graphical Fusion Calculus, I. Lanese and U.
  Montanari, Proceedings of CoMeta Computational
  Metamodels Final Workshop, ENTCS 104
• Mapping Fusion and Synchronized Hyperedge
  Replacement into Logic Programming, I. Lanese
  and U. Montanari, under review for publication in
  a special issue of TPLP

37
Bibliography (new)

• Synchronization algebras with mobility for graph
  transformations, I. Lanese and U. Montanari,
  Proceedings FGUC 2004, ENTCS, to appear
• Synchronized Hyperedge Replacement for
  heterogeneus systems, I. Lanese and E. Tuosto,
  submitted to COORDINATION 2005
• New insights on architectural connectors, R.
  Bruni, J.L. Fiadeiro, I. Lanese, A. Lopes and U.
  Montanari, Proceedings IFIP TCS 2004, Kluwer
  Academics
• Normal forms for stateless connectors, R.
  Bruni, I. Lanese and U. Montanari, submitted to
  CALCO 2005

38
End of talk
Thanks
Questions?
39
Observational and abstract semantics

• Analyzing observational and abstract semantics of
  models
• Useful for abstracting details
• Important to analyze relationships with
  compositionality
• Tile Model is well suited for this kind of
  analysis
• Existing criteria for compositionality too
  restrictive for GC applications

40
Hybrid models

• Plugging features from a model into another one
• Expressiveness and properties of new models must
  be analyzed
• Ongoing work
• Logic programming with restriction
• Concurrent Fusion Calculus
• New ideas
• Non Milner Fusion Calculus
• Non Hoare logic programming

41
Algebra of synchronization strategies

• Synchronization algebras
• Categorize synchronization strategies
• Extending with mobility, name-handling,
• Apply to existing models to make them parametric
  w.r.t. synchronization strategies
• Studying models where different synchronization
  strategies coexist
• SHR good for first analysis