Download presentation
Presentation is loading. Please wait.
1
SHReQ: Coordinating Application Level QoS Speaker: Ivan Lanese Dipartimento di Informatica Università di Pisa Authors: Dan Hirsch & Emilio Tuosto Dipartimento di Informatica Università di Pisa SEFM 2005, Koblenz, Germany, 6-9 September 2005
2
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Roadmap Motivations Background: c-semiring Background: SHR SHReQ An application Final remarks
3
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Roadmap Motivations Background: c-semiring Background: SHR SHReQ An application Final remarks
4
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Global computing Programming global systems is hard because: absence of centralised control client-server not enough: P2P different administrative domains (security) interoperability different platforms, devices (e.g. PDAs, laptops, mobile phones...) mobility network awareness applications are location dependent different locations have different features
5
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Service oriented computing (SOC) Applications are made by gluing services interconnected but independent (local choices, independently built) Interactions governed by programmable coordination policies Services are searched and bound... offline can search/bind be dynamic and at run-time? Search and bind depend on application level QoS not low-level performance (e.g., throughput, response time) but application-related price of services payment mode data available in a given format
6
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Our approach SOC modelling is not just normal programming + primitives for mobility or synchronization Lifting QoS issues to application level proof techniques and tools are also needed First steps extending Klaim [see biblio] We present SHReQ, an (hyper)graph model which exploits c-semiring for expressing application level QoS coordinating activities by synchronisation on c-semiring values
7
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Roadmap Motivations Background: c-semiring Background: SHR SHReQ An application Final remarks
8
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Constraint semirings C-semirings for abstracting application level QoS where A is a set (containing 0 and 1), +, ⋆ : A × A A
9
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Constraint semirings + ⋆ x + y = y + x x ⋆ y = y ⋆ x (x + y) + z = x + (y + z) (x ⋆ y) ⋆ z = x ⋆ (y ⋆ z) x + 0 = x x ⋆ 0 = 0 x + 1 = 1 x ⋆ 1 = x x + x = x (x + y) ⋆ z = (x ⋆ z) + (y ⋆ z) Implicit partial order: a ≤ b a + b = b “b is better than a”
10
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Example: priority semiring P = N is the set of natural numbers with infinity ≤ is the standard order on N
11
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Example: broadcast semiring B = A is a set of actions A = {a | a A} are the coactions a ⋆ b obeys the axioms for 0 and 1 and a ⋆ a = a a ⋆ b = otherwise a + b obeys the axioms for 0 and 1 and a + a = a a + b = otherwise
12
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Other semirings (boolean): Availability (optimization): Price, propagation delay (max/min): Bandwidth (probabilistic): Performance and rates with N set (set-based): Capabilities and access rights
13
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) A key property Cartesian product of c-semirings is a c- semiring Allows to use multiple QoS dimensions at the same time
14
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Roadmap Motivations Background: c-semiring Background: SHR SHReQ An application Final remarks
15
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) The hypergraph model We aim at modeling new non-functional computational phenomena The metaphor is: “Global computers as hypergraphs” “Global computing as SHR” Components are represented by (hyper)edges Systems are bunches of edges connected through common nodes some nodes are declared as local Computing means to synchronously rewrite edges according to a synchronisation policy
16
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) The SHR approach Productions to describe the behaviour of single edges local effects edges rewritten into generic graphs constraints on surrounding nodes Global constraint-solving algorithm to derive transitions to select which productions can be applied allows to define complex transformations Finally transitions are applied
17
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Edge Replacement Systems R 1 2 3 4 L 1 2 3 4 H A production describes how the edge L is transformed into the graph R
18
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Edge Replacement Systems A production describes how the edge L is transformed into the graph R Many productions can be applied concurrently R R’ 1 2 3 4 1 2 3 L L’ 1 2 3 4 1 2 3 H
19
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Adding synchronization We associate actions to nodes attached to the rewritten edges Actions performed on the same node are composed according to a synchronization algebra A transition is allowed iff the synchronization constraints expressed by actions are satisfied (i.e., composition is defined) Many synchronization models are possible (point-to-point, broadcast, …)
20
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Adding mobility We use synchronization algebras with mobility [see biblio] Each action is equipped with a tuple of node references During synchronization we compute a fusion among nodes merging corresponding nodes from synchronized actions mobility in the Fusion Calculus style apply the fusion to the graph and to the references of the resulting action
21
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Roadmap Motivations Background: c-semiring Background: SHR SHReQ An application Final remarks
22
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) From SHR to SHReQ Graphs have weights associated to nodes values of the semiring Productions have an applicability condition the minimum (w.r.t. ≤) value of resources required on each node Actions are values of the semiring ⋆ is action composition
23
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Sample production a x x L y:v y RR z 1 Rewritten edge Final graph
24
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Sample production L a R x x y:v y R z 1 Action Applicability condition Action sending node
25
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) SHReQ synchronization Semiring set has the following structure NoSync Sync Fin 0 1 Synchronization errors Completed synchronizations
26
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Quasiproductions Obtained from productions by renaming nodes from the LHS If nodes are merged, + is used to compute the resulting applicability condition
27
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Synchronizing rules We select a quasiproduction matching each edge same label and attachment nodes we check that the applicability condition is satisfied We compute the product of the actions performed on each node we check that it is not in NoSync On local nodes we check that the resulting action is in Fin We compute the substitution on nodes we check that no old nodes are merged
28
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) The result Each edge is replaced by the RHS of the corresponding quasiproduction The new values on nodes are the result of the synchronization for old nodes 1 for new nodes We apply the induced substitution since no old nodes are merged, values from old nodes can be used We hide nodes if the environment is not aware of them not free before and not extruded as names in -calculus
29
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) There is some math behind Graphs as terms in an algebra Productions and transitions as labelled rules Inference rules to derive transitions from productions
30
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Roadmap Motivations Background: c-semiring Background: SHR SHReQ An application Final remarks
31
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) TRMCS case study We apply SHReQ to Teleservice and Remote Medical Care System (TRMCS) studied by Parco Scientifico e Tecnologico d’Abruzzo and University of L’Aquila Hierarchical system with a server S, routers R connected to S and users U connected to routers Functionalities users send requests for assistance (remote or ambulance) to routers routers forward alarms to server server provides the required assistance Aimed at showing the rewriting mechanism for the use of c-semirings for multidimensional QoS see the bibliography
32
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Modelling the system The cartesian product of priority and broadcast is used All nodes are local (result must be a coaction) S R2R2 R1R1 U2U2 U3U3 U1U1
33
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) User productions UnUn (amb,n) wa U n x x 1 UnUn x x (amb,n) ua U n x x:(amb,n) z
34
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Router productions R (amb,∞) ra R x x (amb,∞) y y ra R (amb,∞) R x x y:(amb,0) y
35
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Server production x (amb,∞) y S w (rem,∞)(amb,∞) xy S w
36
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Sample computation S R2R2 R1R1 U2U2 U3U3 U1U1 (amb,1)(amb,3) (amb, ∞ ) (rem, ∞ ) x y w r s
37
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Sample computation S R2R2 R1R1 U2U2 wa U 3 wa U 1 x y w r s (amb,1) (amb, ∞ ) 1 (rem, ∞ )
38
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Sample computation S R2R2 R1R1 U2U2 wa U 3 wa U 1 (amb, ∞ ) (rem, ∞ ) x y w r s (amb,1)
39
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Sample computation S R2R2 R1R1 U2U2 U3U3 ua U 1 x y w r s
40
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Roadmap Motivations Background: c-semiring Background: SHR SHReQ An application Final remarks
41
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Final remarks We have presented the idea and an application of SHReQ The main features are general multiparty synchronisations based on c-semirings support for application level QoS Surprisingly, SHReQ fits the design principles for a good calculus from Milner [see biblio]
42
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Future work We plan to develop formal methods based on SHReQ for specifying distributed applications and their non-functional requirements Hopefully, develop verification techniques (e.g., model checking) on SHReQ logic for SHReQ already exists [see biblio]
43
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Bibliography On extending KLAIM with QoS: “A formal basis for reasoning on programmable QoS”, De Nicola et al., International Symposium on Verification, LNCS 2772 “A basic calculus for modelling service level agreements”, De Nicola et al., Coordination 2005, LNCS 3454 On c-semiring: “Semiring-based constraint satisfaction and optimization”, Bistarelli, Montanari, Rossi, JACM 44(2) On SHR: “A model of distributed systems based on graph rewriting”, Degano, Montanari, JACM 34 “Reconfiguration of software architecture styles with name mobility”, Hirsch, Inverardi, Montanari, Coordination 2000, LNCS 1906 Ph.D. thesis of Dan Hirsch and Emilio Tuosto
44
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese) Bibliography On synchronization algebras with mobility: “Synchronization algebras with mobility for graph transformations”, Lanese, Montanari, FGUC 2004, ENTCS, to appear On verification techniques for QoS: “A logic for graphs with QoS”, Ferrari, Lluch- Lafuente, VODCA 2004, ENTCS, to appear On a logic for SHReQ: “A logic for application level QoS”, Hirsch, Lluch- Lafuente, Tuosto, submitted to QAPL 2005 On design principles for calculi: “Process constructors and interpretations”, Milner, Information Processing, Elsevier
45
SEFM 2005, 5-9 September, Koblenz, Germany Hirsch, Tuosto (presented by Lanese)
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.