Availability Simulation of Peer-to-Peer Architectural Styles Simon Giesecke, Timo Warns, Wilhelm Hasselbring Referee: Timo Warns
Motivation Evaluation of availability of P2P services Specifics of P2P context impacting availability Failure distribution of peers Means of handling failures Dynamic architecture / topology How to integrate these aspects? Focus: Architectural Style 2 Availability Simulation of Peer-to-Peer Architectural Styles
Approach Conceptual framework P2P styles Architectural Style Fault P2P architectures Characteristics P2P systems Architecture Evaluation by simulation Real-World Simulated “most real-world systems are System System too complex to allow realistic models to be evaluated analytically” Predicted Actual Availability Law and Kelton, 2000 Availability Flexible 3 Availability Simulation of Peer-to-Peer Architectural Styles
Peer-to-Peer Styles Classification scheme Server Peer Type of decentralization Peer Decentralized, hybrid, super-peer Peer Peer Type of communication Direct, Indirect, Mediated Peer Peer Peer Structural Characteristics Super Super Ring, Tree, Small-World Network Peer Peer Rules for evolution Peer Peer Super Joining / leaving of peers Peer Peer No formalisation yet Peer 4 Availability Simulation of Peer-to-Peer Architectural Styles
Architecture Description Model Graph-based formalism A = (N, C, ν, λ, τ) N, C – Sets of nodes and connections ν: C →{{n 1 , n 2 } | n 1 ≠ n 2 and n 1 , n 2 in N} – Node function λ: N → L – Labelling function L is a set of node labels (e.g., “Peer”, “Server”, ...) τ: T → NC T – Time mapping τ describes evolution over time E.g., peer p participates at system from t n to t m => p is in image of τ for t in [t n , t m [ 5 Availability Simulation of Peer-to-Peer Architectural Styles
Example Description Model [t 0 , t 1 [ N = {p 1 , ..., p 4 } Peer p 1 c 1 c 2 C = {c 1 , ..., c 5 } c 3 Peer Peer p 3 p 2 λ(n) = Peer for all n in N [t 1 , t 2 [ Peer Peer τ: ν: p 1 c 1 p 4 c 2 c 4 c ν(c) T NC T c 3 Peer Peer c 1 {p 1 , p 3 } p 3 p 2 [t 0 , t 1 [ p 1 , ..., p 3 , c 1 , ..., c 3 c 2 {p 1 , p 2 } [t 1 , t 2 [ p 1 , ..., p 4 , c 1 , ..., c 4 c 5 c 3 {p 2 , p 3 } [t 2 , t 3 [ [t 2 , t 3 [ p 1 , ..., p 4 , c 1 , ..., c 5 Peer Peer p 1 c 4 {p 3 , p 4 } c 1 p 4 c 2 c 4 c 5 {p 1 , p 4 } c 3 Peer Peer p 3 p 2 6 Availability Simulation of Peer-to-Peer Architectural Styles
Simulation Prototype of simulator 4,50 ROWA 4,25 Based on graph formalism 4,00 Majority Consensus 3,75 Peer model 3,50 3,25 3,00 Relative Change in % Derived from real-world 2,75 2,50 2,25 system 2,00 1,75 Enhanced by classic 1,50 1,25 replication strategies 1,00 0,75 Evaluation of availability of 0,50 0,25 0,00 replicated resources -0,25 1 2 3 4 5 6 7 8 9 10 11 12 Scenario Class 7 Availability Simulation of Peer-to-Peer Architectural Styles
Conclusions Conceptual framework Evaluation of availability of P2P services Architectural styles, architectures, systems Classification scheme for architectural styles Description model for P2P architectures Simulator prototype 8 Availability Simulation of Peer-to-Peer Architectural Styles
Future Work Formalisation of architectural styles Graph grammars? Benefit: Automated creation of architectures Formalisation of peer model Add peer model to input for simulation UML? Development of improved simulator Prototype used manually created architectures and one fixed peer model 9 Availability Simulation of Peer-to-Peer Architectural Styles
Recommend
More recommend